Cassava Starch Pearls as a Desiccant for Drying Ethanol - Industrial

Jun 14, 2011 - Weldon School of Biomedical Engineering, 206 S Martin Jischke Drive, Purdue University, West Lafayette, Indiana 47907-2022, United Stat...
9 downloads 10 Views 2MB Size
ARTICLE pubs.acs.org/IECR

Cassava Starch Pearls as a Desiccant for Drying Ethanol Youngmi Kim,†,‡ Rick Hendrickson,†,‡ Nathan Mosier,†,‡ Ahmad Hilaly,|| and Michael R. Ladisch*,†,‡,§ †

Laboratory of Renewable Resources Engineering, 500 Central Drive, Purdue University, West Lafayette, Indiana, United States Department of Agricultural and Biological Engineering, 225 South University Street, Purdue University, West Lafayette, Indiana, United States § Weldon School of Biomedical Engineering, 206 S Martin Jischke Drive, Purdue University, West Lafayette, Indiana 47907-2022, United States Archer Daniels Midland Co., James R. Randall Research Center, 1001 Brush College Road, Decatur, Illinois 62521, United States

)



ABSTRACT: The fuel ethanol industry uses corn grits packed in fixed bed adsorption towers to dry hydrous ethanol vapors in an energy efficient manner. Spherical micropearl cassava starch exhibit a higher adsorption capacity than corn grits of the same size and may be a viable replacement for ground corn. Adsorption equilibrium curves, BET surface area measurements, and SEM images provide an explanation for the enhanced performance of cassava micropearls based on particle architecture and the surface area available to water molecules. The SEM images show that the micropearls form a coreshell structure with pregel starch acting as the scaffold that holds starch granules in an outer layer. This layer determines the BET surface area and the measured equilibrium adsorption capacity. The coreshell microstructure results in a shortened diffusion path-length and enhanced adsorption rates. These microstructural and operational characteristics provide a template for microfabrication of enhanced capacity starch based spherical adsorbents that could replace ground corn for the drying of ethanol.

1. INTRODUCTION The combination of a conventional distillation process with an appropriate adsorption system to break the waterethanol azeotrope and dry ethanol reduces overall energy requirements in producing fuel-grade ethanol.13 Corn grits, a starch-based adsorbent, have been proven as an energy-efficient desiccant to remove water from alcohols.49 Its first industrial use was a result of joint effort between Purdue University and Archer Daniels Midland (ADM) in 1984. Unlike commercial inorganic adsorbents such as molecular sieves or silica gels, corn grits and cassava are biologically based, biodegradable, nontoxic and derived from renewable biomass. Dehydration of 92 to 93% (by weight) ethanol to yield fuelgrade ethanol (99.5% ethanol by weight) on an industrial scale is done by using a multiple tower fixed bed adsorption system of which one is under adsorption mode while at least one other is in regeneration mode. Water is selectively removed from a hydrous ethanol vapor by corn grits during the adsorption cycle and desorbed from the corn grits by hot CO2 gas (heated to about 96 °C) that is counter-currently passed through the bed during regeneration. The heat of water adsorption released during the feed cycle is stored in the bed and used to help dry the bed during the regeneration cycle. There have been extensive studies on the equilibrium and kinetic aspects of water removal by various types of starch adsorbents on a laboratory scale. The starch adsorbents studied include native corn grits,10,11 modified corn grits,12 corn meal,2,5,6,1315 wheat flour,16 and synthesized starch-based adsorbents.17 The mechanism of water adsorption and equilibrium kinetics and chemistry of maize-derived starch materials, such as corn grits and corn meal, are discussed in many papers. However, cassava starch has been investigated only recently.18,19 r 2011 American Chemical Society

Cassava starch is the fourth largest commercial source of starch after corn, wheat, and potato20 and is abundantly produced in various tropical regions. The main component is amylopectin which comprises nearly 80% of the total starch.21,22 The world production of cassava starch is approximately 250 million tons per year.23 Spherical pellets of cassava starch, which are generally referred to as “tapioca pearls”, are commercially available as food additives. Preparation of pearl shaped cassava starch granules generally involves the following steps:24 granulation of wetted starch into beads, stirring the beads on a hot plate, drying the resulting pearls at 4060 °C in a stream of hot air to approximately 10% moisture, and cooling and packing according to the pearl size. The spherical shape occurs when small starch particles adhere to one another.25 The process involves heat-moisture treatment that leads to gelatinization that causes structural disruption of starch. Upon cooling, the gelatinized starch is retrograded, i.e., recrystallized. Commercial tapioca pearls contain approximately 60% gelatinized starch.26 Because cassava pearls have a spherical shape, a hard texture, and are commercially available in a narrow range of defined particle diameters, they have potential as a useful replacement for corn grits as a desiccant material. Their application as adsorbents for water first reported by Carmo et al.,18 who showed the effectiveness of cassava starch in adsorbing water from various liquid-phase alcoholwater mixtures. Adsorptive ethanol drying in industry is usually carried out under vapor phase conditions. Received: February 16, 2011 Accepted: June 14, 2011 Revised: June 13, 2011 Published: June 14, 2011 8678

dx.doi.org/10.1021/ie2003297 | Ind. Eng. Chem. Res. 2011, 50, 8678–8685

Industrial & Engineering Chemistry Research

ARTICLE

Figure 1. Process schematic for adsorption apparatus.

The goal of this paper is to report the potential use of cassava starch in the form of spherical pellets (tapioca pearls) as a desiccant for dehydration of ethanol under vapor phase conditions. Spherical cassava pearls with varying diameters and corn grits were tested and compared in terms of adsorption capacity and selectivity for water under identical operating conditions. Fundamental understanding of the properties and structural characteristics of the spherically shaped starch pellets as an ethanol drying desiccant is expected to benefit the design of synthetic starch-based adsorbent systems.

2. MATERIALS AND METHODS 2.1. Materials. Cassava (tapioca) pearls having a nominal average particle diameter of 0.6 mm were purchased from Industria Agro Commercial Cassava S/A (Rio do Sul-SC, Brazil). The tapioca pearls were further sieved into fractions of three different mean particle diameters: 2 mm (810 mesh), 1.0 mm (1820 mesh), and 0.5 mm (3540 mesh). Corn grits having a mean diameter of 1.7 mm were provided by ADM. The aqueous feed solution was prepared by mixing 200 proof ethanol (Pharmco, Brookfield, CT) with deionized water to give various ethanol/water mixtures in the range of 8897% by weight ethanol. All other chemicals were purchased from Sigma-Aldrich (St. Louis, MO) unless specified otherwise. 2.2. Methods. Scanning Electron Microscopy (SEM). Surface images of the cassava pearl adsorbents and corn grits were taken using scanning electron microscopy (SEM, Model JEOL JSM840, JEOL USA Inc., Peabody, MA) at the Life Science Microscopy Facility at Purdue University. Cross-sectional SEM images were also taken from cassava pearls fractured by a small hammer. Samples for SEM imaging were prepared by mounting on aluminum stubs using double-coated tape. Excess material was gently blown off, and the sample was sputter coated with AuPd in the presence of argon gas using a Hummer I sputter coater

(Technics Inc., Alexandria, VA) prior to imaging. All images were taken at 5 kV. Brunauer, Emmett, and Teller (BET) Surface Area Measurement. BET surface areas of tapioca adsorbents samples with three different particle diameters were measured using a Micromeritics TriStar 3000 at Micromeritics Analytical Services (Norcross, GA). Adsorption/Desorption. The system apparatus is shown in Figure 1. The adsorption column was constructed from two glass columns (50  1200 mm and 50  600 mm Pyrex Glass Jacketed Chromatography Column, Ace Glass, Vineland, NJ) connected to each other resulting in a total length of 180 cm (6 feet). For adsorption equilibrium runs, a 1 ft long mesh screen basket packed with a known particle diameter of tapioca pearls was inserted into the 6 ft column, and the space below the basket was left empty. The dry weight of tapioca pearls or corn grits packed into each mesh screen basket was approximately 350 and 320 g, respectively. Breakthrough and desorption characteristics of tapioca adsorbents were determined from a full 6 ft bed packed with 2.5 kg dry pearls or corn grits. Initially, the adsorption bed was prepared for experimentation by drying the packed adsorbents for 4 h at a regenerant gas flow rate of 7.5 L/min and then overnight at 1.75 L/min using 105 °C dry CO2 gas in a down flow direction. The temperature of the column filled with adsorbents was kept at 90 °C by circulation of hot water through the water-jacket. The adsorption column was also insulated using a 2 in.  1/8 in. thick ceramic fiber insulation strip (McMaster-CARR, Elmhurst, IN) to minimize heat loss during experiments. During adsorption cycles, ethanol/water vapor mixture of a known concentration at 110 °C was passed through the packed bed of adsorbents at a vapor superficial velocity of 0.2 m/s until the packed adsorbents reached equilibrium. The “dry” product stream during the feed cycle was condensed in a heat exchanger (HE2) with chilled water. Product composition was monitored 8679

dx.doi.org/10.1021/ie2003297 |Ind. Eng. Chem. Res. 2011, 50, 8678–8685

Industrial & Engineering Chemistry Research

ARTICLE

Figure 2. Scanning Electron Microscopy (SEM) images of cassava starch pearls and corn grits. (A) 1 mm cassava pearl particle; (B) surface of 2 mm cassava pearl particle, heavily populated with starch granules; (C) surface of 1 mm cassava pearl particle with (a) loose dry starch granules and (b) patches of gelatinized starch; (D) surface of 0.5 mm cassava pearl particle, showing extensively gelatinized starch particles; (E) (F) inside of 1 mm cassava pearl particle, showing highly gelatinized internal structure and porous structures near the surface of particles; (G) (H) surface of corn grit particle consisting of (c) smooth vitreous regions with starch granules tightly embedded in a honeycomb like structure and (d) opaque regions of loosely packed starch particles.

online through a density meter (Promass 83, Endress+Hauser, Greenwood, IN) with an accuracy to two decimal places of % wt ethanol to ensure that the packed adsorbents reached equilibrium. The system was regarded to reach equilibrium when there was less than 0.1% change in the product composition over 5 min. Upon completion of the feed cycle, the bed was regenerated with hot CO2 (105 °C) flowing in the countercurrent direction to the feed flow during adsorption at a superficial velocity of 5.6 m/min. Regenerant condensate was collected through three cold traps immersed in buckets filled with dry ice. The regenerant condensate collected in the traps was carefully weighed, and the

composition was analyzed by GC. Bed regeneration was done until there was no increase in regenerant condensate weight. Water and ethanol vapors present in the void space of the bed at the time of regeneration were subtracted from the regenerant condensate. The adsorption capacity of tapioca pearls at equilibrium was calculated and expressed as mg water adsorbed per g of dry tapioca pearls packed in the system. Duplicate runs were made for all experiments unless otherwise noted. After each run, the bed was kept in regeneration mode by flowing 105 °C dry CO2 at 1.75 L/min for at least 12 h until the next run was made. Regenerant condensate collected during the 12 h regeneration period was also collected and analyzed. The dryness of the system 8680

dx.doi.org/10.1021/ie2003297 |Ind. Eng. Chem. Res. 2011, 50, 8678–8685

Industrial & Engineering Chemistry Research for the next run was confirmed by measuring humidity of the regeneration off-gas using a hygrometer (Humidity and temperature probe HMP368, Vaisala Oyj, Vantaa, Finland). All runs given in this study represent data obtained from preequilibrated adsorbents for which at least two preruns were made. None of the data presented in this paper were obtained with fresh adsorbents. Data Monitoring. Temperature, pressure drop across the bed, moisture content, flow rate of dry CO2, and mass of feed and product were monitored through a LabVIEW compact field point data acquisition system (National Instruments Co., Austin, TX). The masses of the feed reservoir and product collection reservoir were measured continuously using digital balances. Temperatures of the system were also measured and monitored by thermocouples and temperature transmitters. The flow rate of dry CO2 was monitored and controlled using a mass flow meter/ control (Brooks Model 5851S, Brooks Instruments, Hatfield, PA). The communication module was a LabVIEW Real-Time/ Ethernet Network Module (cFP-2000, National Instruments Co., Austin, TX). Software to acquire and process signal data was prepared by VI Engineering (Indianapolis, IN). Gas Chromatography. Ethanol concentration of feed, product, and regenerant condensate was measured using a gas chromatography system (Varian 3400 Gas Chromatography, Varian Inc., Palo Alto, CA) with a HayeSep P column (80  1/800 SS, 60/80 mesh, Hayes Separations, Bandera, TX), integrator (Agilent 3395 Integrator, Agilent Technologies, Palo Alto, CA), and compressed gas tank containing grade 5 helium as a carrier gas. The amount of sample injected was 2 μL. Temperatures of the column, injection, and detector were 120, 200, 200 °C, respectively. Flow rate was 30 mL/min. Each sample analysis was completed in 5 min. A calibration curve of the system was determined with standard ethanol/water mixtures with varying concentration of water. Sample composition was determined by peak area.

3. RESULTS AND DISCUSSION 3.1. Surface Characteristics of Cassava Pearls. Cassava tapioca pearls are spherical (Figure 2(A)) and, unlike irregularly shaped corn grits, are free of dust. Micropearls roll when poured into columns to form well-packed beds. The surfaces of the pearls are populated with 510 μm starch particles interspersed with smooth regions of gelatinized and aggregated starch (Figure 2(B), (C), (D)). As particle size decreases from 2 to 1 to 0.5 mm, the number of distinct starch particles on the surfaces of the pearls decreases and gelatinized area increases (compare parts (B), (C), and (D) of Figure 2). The inside of the pearl is gelatinized and has little porosity (Figure 2(E), (F)). Hence, pore structures that would provide surface area for adsorption are on the surface. The overall structure is coreshell with pregel (gelatinized starch) serving as the scaffold that holds starch particles to the outer layer of the particle. The outer layer of starch should account for most of the surface area and adsorption capacity, while the core has limited porosity and capacity. While both cassava pearls and corn grits have surfaces that consist of tightly packed spherical starch particles,27,28 the starch particles in corn grits are embedded in a honeycomb matrix12 in the horny endosperm or loosely held in the floury endosperm, rather than fixed by gelatinization, as is the case for cassava pearls (see Figure 2(G) and (H)). The ratio of horny to floury endosperm is approximately 2:1, thus resulting in dust

ARTICLE

Table 1. BET (Brunauer-Emmett-Teller) Surface Areas of Cassava Micropearl Adsorbentsa BET surface area (m2/g)

a

dp = 2.0 mm

0.418 ( 0.002

dp = 1.0 mm dp = 0.5 mm

0.563 ( 0.003 0.516 ( 0.002

dp = diameter of cassava pearl adsorbent.

formation.12 Only a fraction of the total sorption sites of corn grits is accessible during water sorption process.10,27 Water adsorption mainly occurs onto the starch granules, where water molecules have easy access and swell the starch matrix upon adsorption.29 Amylase enzyme may increase porosity and surface area of grits resulting in an increase of operational adsorption capacity despite the lower mass of starch per volume of the modified, highly porous corn grits.12,28,29 Unlike the corn grit adsorbents, which do not undergo any heat treatment, cassava pearls are precooked and dried during the manufacturing process resulting in a gelatinized/retrograded core. While gelatinization is a process of structural disruption (reduced crystallinity) of starch granules, resulting in increased water sorption capacity of the starch, cooling of the gelatinized starch leads to retrogradation. During retrogradation gelatinized starch regains its crystalline form through hydrogen bonding between realigned molecules.30,31 Thus, gelatinization followed by retrogradation leads to increased crystallinity and reduced water adsorption capacity of starch granules due to loss of available hydroxyl group adsorption sites on the starch surface.28 Since the core of the pearls has a limited water adsorption capacity, the equilibrium adsorption capacity is strongly dependent on the outer surface area of the pearls. The relation between surface area and adsorption capacity of cassava pearl adsorbents was, therefore, further investigated through measurement of BET adsorption isotherms. 3.2. BET Surface Area of Cassava Pearls. The pearls in the 0.52.0 mm particle diameter range have 0.42-0.56 m2/g BET surface areas (Table 1). Starch particles with a diameter of 10 μm, in comparison, have surface areas of 0.72 m2/g17. Considering that the pearls have no significant internal pores as shown in the SEM images, the BET surface area of the cassava pearls is due to starch particles that are fixed to the external surfaces of the cassava micropearls. The surface area of the 1 mm particle diameter pearls at 0.56 m2/g was 35% higher than that of 2 mm pearls at 0.418 m2/g (Table 1). The SEM images showed that smooth and gelatinized regions became more dominant as the particle size approached the size of its pregelatinized core. Hence the BET surface area of 0.5 mm cassava micropearl, while 23% larger than 2 mm particles, was 10% less than that of 1 mm beads. While representing a qualitative metric, the SEM images in Figures 2 show a larger extent of agglomerated starch particles for the 0.5 mm pearls than the 1 mm diameter particles, thus helping to explain the lower surface area, despite the smaller particle diameter. Confirmation of the surface area effect was obtained by measurements of equilibrium adsorption capacity. 3.3. Equilibrium Adsorption Capacity of Cassava PearlShaped Adsorbents. The equilibrium adsorption isotherm for the water-starch system is generally well described by type II isotherm for a low relative humidity range.10,11 The type II adsorption isotherm represents formation of a monolayer followed by unlimited multilayer adsorption as the concentration of 8681

dx.doi.org/10.1021/ie2003297 |Ind. Eng. Chem. Res. 2011, 50, 8678–8685

Industrial & Engineering Chemistry Research

ARTICLE

Figure 3. Linear adsorption equilibrium curves for cassava pearls and corn grits adsorbents based on duplicate measurements. Isotherms generated by vaporizing a liquid feed at the indicated concentrations, contacting the vapor with the adsorbents, and measuring the amount of water adsorbed from the vapor phase onto the cassava pearls or corn grits. Compositions represent inlet concentrations passed over the adsorbents. Adsorption equilibria was indicated when inlet and outlet compositions were the same.

adsorbate increases. Since distillation at atmospheric conditions gives 85% to 92% (w/w) ethanol, with a maximum of 95.6% (i.e., the azeotrope), the resulting vapor has a low (molar) concentration of water in an ethanol vapor, resulting in monolayer adsorption, and a linear isotherm that follows the equation (similar to Henry’s Law)2,5,911 q ¼ K3C

ð1Þ

in which q is the mass of water adsorbed per mass of adsorbents (mg water/g dry adsorbent), C is the water vapor concentration (mg water/cm3), and K is the linear adsorption equilibrium constant (cm3/g dry adsorbent). Both starch-based adsorbents, cassava pearls and corn grits, exhibit linear behavior at concentrations of 0.04 to 0.15 mg water/mL ethanol (Figure 3). This range corresponds to 88 to 97% w/w ethanol. The constant K was determined from linear regression of the data points with Pearson’s coefficient of regression (r2) of 0.92 or greater. The resulting values were consistent with the qualitative observations made by SEM. The 1 mm pearls had 40% higher equilibrium capacity (and K value) than the 2 mm pearls, indicating a specific surface area effect. The K value for the smallest particle tested (0.5 mm) was 5% lower than that of the 1 mm particles, which was consistent with the larger extent of agglomerated starch spheres/particles characteristic of the 0.5 mm pearls. The equilibrium constant K also correlated linearly with the measured BET specific surface area (Figure 4). The linear regression for the cassava pearls gave a 0.96 Pearson’s coefficient of regression (r2). Adsorption equilibrium capacity and the number of adsorption sites of cassava pearls are, therefore, shown to be dependent on the specific surface area, rather than just the mass of starch adsorbent. Water molecules have limited access to the inside of the pearls due to the low porosity and the gelatinized/ retrograded core. Equilibrium water loading for corn grits having an average diameter of 1.7 mm and a BET surface area of 0.22 m2/g28 was lower than all of the cassava pearls tested (Figure 3 and Figure 4). The lower water sorption capacity of corn grits follows from the lower surface area compared to that of an equivalent diameter of cassava pearls. Lower starch content of grits relative to the pearls

Figure 4. A plot of constant K (linear adsorption equilibrium constant) versus measured BET surface area of cassava pearls per unit mass. Numbers are average of duplicate runs.

may also lead to lower capacity. While the cassava pearls are entirely made of cassava starch flour, the corn grits contain other components, such as protein and oil, which are less effective than starch in terms of adsorbing water molecules.32 Approximately 4050% of endosperm protein in corn is known to be zein, which exhibits poor solubility in water due to its content of nonpolar amino acids.33,34 However, the impact of differences in composition is likely to be secondary to the effect of accessible starch in determining capacity. Only a small portion of the potential adsorption sites of starch materials are utilized during adsorption as most sites are embedded in the interior of the starch matrix. Water must diffuse into the matrix in order for adsorption to occur.27,35 Based on measurements of BET surface area and experimental measurement of adsorption capacities, particulate starch alone which has characteristic diameters of 10 to 100 μm would be the preferred adsorbent giving the highest capacity. While equilibria would be favored, the pressure drop of 10 to 100 μm diameter starch particles in a packed bed is very large. Hence cassava pearl is an attractive option providing capacity, low pressure drop, and robustness. The center core of cassava pearls serves as a supporting scaffold for immobilized starch particulates on the external surface, providing mechanical stability to the individual particle. The size of the center may be controlled during the production process to give different particle diameters. Increased accessible surface area is achieved by manipulating the coating and immobilization procedures to give a thicker shell of starch agglomerates deposited onto the particle’s core. When combined, these characteristics resulted in a stable spherical particle with higher operational adsorption capacity than starch or corn grits, as was shown in experiments carried out with packed beds of these materials. 3.4. Operational Adsorption Capacity of Cassava Pearl in Packed Beds. Separation of water from alcohol vapor in a fixed bed adsorber is a kinetically controlled process in which water adsorption occurs 1001000 times more rapidly than ethanol.4,5,27,35 Even though the diffusion rate of water through starch materials is known to be relatively slow,36 contact times of hours to days allow the water molecules to penetrate the intrastarch hydrogen bonds in the starch matrix, causing the structure to swell and open-up, thereby, maximizing the water loaded on the adsorbents.17 A diffusion gradient created through the adsorbents results in kinetically controlled water adsorption if 8682

dx.doi.org/10.1021/ie2003297 |Ind. Eng. Chem. Res. 2011, 50, 8678–8685

Industrial & Engineering Chemistry Research

ARTICLE

contact times are short (minutes to hours), since adsorption is localized to the surface of the starch. Ethanol adsorption becomes significant at contact times of hours to days, thereby, lowering the selectivity of adsorption of water over ethanol.14,37 In a packed-bed adsorption system, contact times are kept short so that there is sufficient time for water to adsorb at a temperature below its normal boiling point but higher than that of ethanol (i.e., 80 to 100 °C). The system is operated so that the adsorption process never reaches equilibrium and is stopped at the breakthrough point to give a dry ethanol of g99.5% wt. The breakthrough time determines the operational capacity of adsorbents with a longer time corresponding to a higher capacity. Adsorption on the surface of starch becomes the dominant factor in determining the kinetic selectivity of water molecules and operational water adsorption capacity. The extent of water adsorption and selectivity typically increases as the particle diameter of adsorbent decreases.14,17,34 Since surface areas increases with decreasing particle size, a higher capacity is achieved with a smaller particle size of adsorbents. Mass transfer is also enhanced by smaller particle size, and this results in rapid equilibrium as well. When a fast equilibrium is established, the breakthrough curve exhibits a sharp front and a long breakthrough time. Breakthrough curves of a 6 ft fixed bed of corn grits having an average diameter of 1.7 mm and cassava pearls of 1 mm and 0.5 mm diameter are presented in Figure 5 for a vapor feed of 93.7% w/w ethanol. The data represent an average of duplicate runs. The x-axis was adjusted so that the graph highlights the adsorption wavefront, rather than showing the entire adsorption curve until it reaches an equilibrium product concentration. Sorption conditions were identical for all breakthrough runs. There was a delay of about 1.5 min between the start of pumping the vapor

Figure 5. Breakthrough curves of cassava pearls (0.5 mm, 1 mm in diameter) and corn grits (1.7 mm diameter) adsorbents. Feed ethanol concentration: 93.7% w/w.

feed to the column and collecting the first drop of condensate. Figure 5 represents an elapsed time from the start of the feed cycle. The breakthrough time was 7.5 min for corn grits and 22 and 24 min for 1 mm and 0.5 mm pearls, respectively. The significantly shorter breakthrough time for corn grits was expected since corn grits have smaller surface areas and lower equilibrium adsorption capacities as compared to those of cassava pearls (Figure 3 and Figure 4). The operational pressure drops in these runs, which were carried out at an average gas velocity of 0.2 m/s, were 0.27 psi/m for cassava micropearls versus 0.57 psi/m for corn grits. The calculated pressure drop per length of the bed by Ergun equation was 0.23 psi/m for the 2 mm particle size cassava pearls, assuming void fraction of 0.35. Operational adsorption capacity at the breakthrough point and equilibrium adsorption capacity of a 6 ft bed of cassava pearls are summarized in Table 2. Selectivity of each adsorbent was compared by a separation factor which is defined as37 R¼

Xw =Xe Yw =Ye

ð2Þ

where Xw and Yw are the mass fractions of water in the sorbed phase and vapor phase at equilibrium, respectively, and Xe and Ye are the corresponding ethanol mass fractions. At breakthrough, water and ethanol sorption capacities were not significantly different for the two particle sizes of pearls. The 0.5 mm pearls gave a slightly higher sorption capacity for water than the 1 mm pearls and is expected since mass transfer is higher with its smaller particle size, although surface characteristics moderate the benefits of smaller particle size, as discussed earlier in this paper. The separation factor at breakthrough was about the same within error limits for both 0.5 mm and 1 mm pearls. At equilibrium, the ethanol loading for 0.5 mm pearls was nearly twice that for 1 mm particles, resulting in a greater separation factor for the 1 mm pearls. Equilibrium adsorption capacity represents a maximum water loading at the given conditions. Although the system reached equilibrium in terms of water sorption, ethanol adsorption might have not reached equilibrium during the 3 h equilibrium measurements as it takes much longer for ethanol to reach equilibrium.5,35 Selectivity is also compromised if adsorbed water further absorbs ethanol, extending the time required for ethanol to reach equilibrium state. The amount of desorbed ethanol was highest for the smallest (0.5 mm) pearls. As shown in the SEM images in Figure 2, the 0.5 mm pearls display a mostly smooth external surface due to the gelatinized starch, a lower BET surface area, and a surface that is more likely to adsorb ethanol. Gelatinized starch has a lower selectivity for water.37 Crawshaw and Hills37 reported that cooked corn grits have a slightly greater total sorption capacity but much less selectivity for water than uncooked grits at equilibrium. Chang et al.5,6 reported a separation factor in a range of 1030 at equilibrium for a bed of cornmeal adsorbent with a granularity

Table 2. Operational and Equilibrium Adsorption Capacities and Separation Factors for Cassava Pearlsa adsorption capacity at breakthrough (mg/g)

a

equilibrium adsorption capacity (mg/g)

water

ethanol

separation factor

water

ethanol

separation factor

cassava pearls dp = 1 mm

15.2

5.4

47.9

25.8

6.8

63.6

cassava pearls dp = 0.5 mm

16.4

5.2

53.0

26.5

12.2

36.5

Numbers represent an average of duplicate runs. 8683

dx.doi.org/10.1021/ie2003297 |Ind. Eng. Chem. Res. 2011, 50, 8678–8685

Industrial & Engineering Chemistry Research

ARTICLE

saturated portion of the bed.2 In this study, the bed regeneration was initiated at the breakthrough point for respective particle diameter cassava pearls. Regeneration effluent composition and desorption profiles are given in Figure 6(A) and (B). The majority of the initial regenerant condensate collected represents the ethanol vapor present in the void fraction of the bed and is swept out in the first 25 min (Figure 6(A)) and then levels off. Water became the major component after 25 min comprising more than 90% of the condensate (Figure 6(B)). While the desorption characteristics of both the 0.5 and 1 mm cassava pearls are similar, the 0.5 mm pearl may initially retain more ethanol than the 1 mm pearl (Figure 6(B)), probably due to retention of ethanol by the aggregated and gelatinized starch regions in the 0.5 mm particles. At the end of 3 h of regeneration, only 75% of the sorbed ethanol was recovered in the regeneration effluent of 0.5 mm pearls, while 92% was recovered for 1 mm pearls. As for water desorption, 93% of the retained water was recovered within 3 h of regeneration for both particle sizes. Desorption rate is significantly slower than adsorption rate and is consistent with the hysteresis effect noted by Hong et al.32 for starch containing adsorbents. Westgate and Ladisch27 provided further explanation for the difference in adsorption and desorption rate based on a trapping mechanism. During adsorption, water molecules penetrate the starch matrix, thus causing starch granules to swell. The swollen starch structure starts to shrink during desorption, thereby trapping some adsorbate in the starch matrix.27,38 Hence, desorbing the trapped adsorbates requires additional time and energy. A cost and energy efficient regeneration strategy has proven to be an important factor in the economics of operation of such a system, and industrial operation of a fixed bed system under thermal swing conditions has been successfully implemented for corn grits.

Figure 6. (A) Ethanol wt % in regenerant condensate; (B) water and ethanol mass desorbed per gram dry adsorbents, does not include ethanol or water recovered from the void volume of the bed or connecting tubing. Mass desorbed is in terms of gram adsorbent in packed bed (dry basis).

of