Adsorption of Glycols, Sugars, and Related Multiple− OH Compounds

3746

Ind. Eng. Chem. Res. 1999, 38, 3746-3753

Adsorption of Glycols, Sugars, and Related Multiple -OH Compounds onto Activated Carbons. 2. Solvent Regeneration Daniel Chinn† and C. Judson King* Department of Chemical Engineering and Lawrence Berkeley National Laboratory, University of California, Berkeley, California 94720

Solvent regeneration with acetone or methanol was explored as a means to recover adsorbed multiple -OH compounds from activated carbons. Batch-desorption experiments with thermogravimetric analysis demonstrated that regeneration can be carried out in situ, but that highenergy sites are more difficult to regenerate. When solvent regeneration is done in a fixed-bed arrangement, a focusing effect occurs where the solute concentration in the effluent rises to three to six times its initial value during the loading stage. Focusing can be maximized using adsorbents with low oxygen content (e3 wt. %), lower volumetric flow rate (3-15 bed volumes/ h), and uniform particle size or particle-size distribution. With methanol as the regenerant, fixed-bed adsorption and regeneration could be operated over several cycles with minimal loss of bed capacity. However, capacity losses occur during acetone regeneration cycles, because acetone is more strongly held on the bed than is methanol. Introduction Part I of this paper1 identified activated carbons as effective adsorbents for removing multiple -OH compounds (glycerol, glycols, sugars) from aqueous solution. The research described herein was concerned with developing an effective method for regenerating the spent carbons in a way that would enable recovery of the adsorbed solute(s). A common regeneration method is to strip adsorbates from carbons using steam or hot inert gases. In the case of steam stripping, the vapor stream can be condensed to a liquid and subsequently distilled. With inert gases, the vapor stream can be cooled to allow condensation of the adsorbed components. The cost for steam regeneration is several cents per pound of regenerated carbon; however, steam stripping is usually restricted to volatile adsorbates with normal boiling points up to 120 °C.2 This method does not appear feasible for the nonvolatile sugars and other high-boiling (Tnbp of 200-300 °C) multiple -OH compounds in the present work. Another nondestructive regeneration method is to change the pH in the proximity of the carbon surface, which causes the adsorption equilibrium between the surface and the ionized form of the adsorbate to become unfavorable.2 Carbon regeneration processes based on pH swing have been used for phenol recovery, where adsorbed phenol (pKa of 9.89) is converted to the phenoxide form using dilute sodium hydroxide.3 A similar method applied to recovery of nonaromatic multiple -OH compounds would be much more problematic, because the pKa values of the hydroxyl groups on glycols and sugars are between 12 and 14.4 * To whom correspondence should be addressed: Provost and Senior Vice President, Academic Affairs, Office of the President, University of California, 1111 Franklin St., 12th Floor, Oakland, CA 94607-5200. Phone: (510) 987-9020. Fax: (510) 987-9209. E-mail: [email protected]. † Current address: Zeneca Ag Products, Western Research Center, 1200 South 47th St., P.O. Box 4023, Richmond, CA 94804. Phone: (510) 231-1129. E-Mail: daniel.chinn@agna. zeneca.com.

A third method of carbon regeneration, and the one investigated in this work, is equilibration of the loaded carbon with an organic solvent that has a greater affinity for the surface than do the adsorbed compounds. The adsorbates are displaced from the surface and dissolved into the solvent. Product recovery from the organic phase can then be effected by conventional distillation, with recycle of the solvent. For reuse of the bed, residual solvent must be rinsed or stripped from the carbon. A substantial amount of research is reported in the literature on solvent regeneration of carbons; however, most of these studies have been confined to phenols and related aromatic compounds.5-8 Acetone and methanol are two of the most common solvents used in solvent regeneration of activated carbons. Both are capable of dissolving a wide range of organic compounds, and have other desirable characteristics of being low-boiling (Tnbp of 56 °C and 65 °C, respectively) and relatively inexpensive ($ 0.39/lb and $0.05/lb, respectively9). Of course, the use of volatile organic solvents in practice requires additional safeguards against flammability hazards, and additional solvent recycle processes to avoid discharge into the environment. The potential for solvent regeneration to recover multiple -OH adsorbates is seen clearly in Figure 1, where measured composite uptakes of ethylene glycol (EG) and glycerol from water and from an organic solvent (acetone or methanol) are shown.4 To account for solution-phase nonidealities, the equilibrium mole fractions are multiplied by the solute activity coefficients.4,10 The data show that the degree of solute adsorption onto the carbon surface from an aqueous solution is substantially greater than that from an organic medium. The results of Figure 1, which are for binary solutions, also pose the important question of whether water entrained within the carbon has any effect on the adsorption/desorption equilibrium. As a gram of carbon will have 0.5-0.7 g of water taken up nonselectively during loading,4 the issue of whether solvent regeneration requires preliminary drying of the carbons had to be addressed.

10.1021/ie990289x CCC: $18.00 © 1999 American Chemical Society Published on Web 09/16/1999

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999 3747

Figure 1. Composite uptake (25 °C) for ethylene glycol (EG) and glycerol from aqueous and organic solvents. All activity coefficients obtained from ref 10. Because of lack of available data, activity coefficient data for EG-acetone at 50 °C were used to determine the solute activities at 25 °C.

Objective and Approach The present work had the dual goals of understanding regeneration equilibria and identifying a regeneration process that enables a favorable recovery of the originally adsorbed solute with reuse of the carbon. Specifically, the experiments related to equilibria were designed to identify the optimum solvent for regeneration, and to study the effect of entrained water in the carbon on the completeness of regeneration. The bench-scale recovery process involved adsorption and solvent regeneration experiments in a fixed-bed apparatus. Volumetric flow rate, carbon type, and particle-size distribution were varied to identify conditions leading to an optimal recovery. EG was the solute in most of the experiments reported herein; however, similar findings have been reported for glycerol, propylene glycol (PG), sorbitol, and sucrose as well.4 Methods and Materials Materials. Chemical Reagents. Purity and vendor information for all reagents are listed elsewhere.4 All solutions were prepared from distilled water that had been passed through a Milli-Q purification system (Millipore Corp.). Adsorbents. The bulk and surface properties of the heat-treated and acid-oxidized F400 (Calgon) carbons are reported in Part I of this work.1 Another adsorbent used in the fixed-bed work was Ambersorb 572 (Rohm and Haas), which is derived from sulfonated styrenedivinylbenzene that has been pyrolyzed. Ambersorb 572 has a surface area of 1100 m2/g, and a bulk elemental composition (by weight) of: 93.31% C, 0.58% H, 3.2% S, and 2.71% O.4 Despite somewhat higher hydrogen and sulfur contents, these numbers are similar to those frequently seen in activated carbons. Methods. Batch-Desorption Experiments. The apparatus and methods for batch equilibrium experiments

are described in Part I of this work.1 In solventregeneration experiments, loaded carbons are equilibrated with pure acetone or methanol at a phase ratio of 10 mL/g carbon. Afterward the carbons were centrifuged as in the batch adsorption experiments, before being subjected to analysis by thermogravimetric analysis (TGA). Solvent regeneration was conducted under either wet or dry conditions. Wet means that the loaded carbon was regenerated immediately after the centrifugation stage. Dry means that after centrifugation, the loaded carbons were vacuum-dried (0.3 atm) at 65 °C for a day before solvent regeneration. Through these relatively mild drying conditions the goal was to evaporate all of the water from the carbon with minimal loss of solute. Fixed-Bed Experiments: Preparation. Before use, 5 to 6 g of dried carbon were prewet by immersion into boiling water for 30 min. The carbon slurry was cooled to 25 °C and then poured into a 1 cm × 30 cm Chromaflex column (Kontes), which was gently tapped to ensure that air bubbles were absent from within the bed. An adjustable plunger contained the solvent line leading into the bed and was positioned in the liquid space of the bed about 5 mm above the carbons. The exit line from the column was connected to a 2211 Superrac fraction collector (LKB Bromma). Bed lengths were from 16 to 19 cm. Fixed-Bed Experiments: Operation. The aqueous feed solution (typically 5 wt. % solute) was driven through the bed using a Masterflex Model 7518-00 pump (ColeParmer). The fraction collector was started once the solution front visibly mixed within the liquid space at the top of the bed. This headspace accounts for about 6% of the total bed volume. Fractions of the column effluent were then collected automatically at fixed time intervals in 20-mL glass scintillation vials. The composition of each liquid fraction was then determined by HPLC, as described in Part I of this work.1 The fraction collector was stopped after breakthrough of the feed solution was complete. The feed was then switched to acetone or methanol, and then the entire procedure described above was repeated. Often, the volumetric flow rate and time intervals for fraction collection were varied between the loading and regeneration stage. Thermogravimetric Analysis. The apparatus and methods for TGA work are described in Part I of this work.1 The heating rates used were either 10 or 15 °C/min. Analysis of Data. TGA Data. All TGA data are reported as derivative plots, from which desorption profiles of the solute are clearly seen. Whereas measuring concentration changes in the bulk solution determines the composite (excess) uptake, integration of the desorption profile from the TGA derivative plot enables determination of the individual (overall) uptake.4 Fixed-Bed Experiments. The elution profiles from the fixed-bed regeneration experiments are plotted as C/Cfeed against the number of bed volumes (BV), where C is the solute concentration (g/mL), Cfeed is the original solute concentration of the aqueous feed (g/mL), and BV is defined as:

BV )

Ft π 2 D L 4 T

(1)

where F is the volumetric flow rate (mL/min); DT is the column diameter (cm); L is the carbon bed length (cm); t is time (min) with t ) 0 defined when solution first

3748

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999

Figure 2. TGA derivative plots for acetone, methanol, and water loaded onto F400/HT carbon. Heating rate: 10 °C/min.

Figure 3. TGA derivative plots for (a) EG-loaded (q ) 23 mg/g) F400/HT carbon, (b) dry methanol (MeOH)-regenerated F400/HT carbon, and (c) bare carbon reference. Heating rate: 10 °C/min.

reaches the top of the bed; and  is the interstitial porosity of the bed, which was determined to be 0.46.4 Results and Discussion Desorption Studies with TGA: Choice of Solvent. Some understanding of how the regenerant solvents interact with the carbon surface is gained from Figure 2, which shows TGA derivatives for carbons immersed in water, methanol, and acetone. The data show that both methanol and water have maximum desorption rates close to the normal boiling points. Acetone, however, remains on the carbon up to 200 °C, which suggests that a substantial fraction of the total acetone is very strongly adsorbed onto the surface. Completeness of Regeneration. To demonstrate how TGA could be used to assess regenerability, Figure 3 shows superimposed derivative plots for loaded and regenerated carbons. Heat-treated F400 carbon was loaded with EG (q ) 23 mg/g), centrifuged, and then vacuum-dried. Part of the carbon was analyzed by TGA, whereas the rest was batch-wise regenerated by methanol (10 mL/g carbon). Before the TGA experiments, the regenerated carbons were first vacuum-dried to remove methanol. The desorption profiles of Figure 3 show that there appears to be a substantial fraction of EG that was not amenable to methanol regeneration. The EG from the regenerated sample exhibits a temperature maximum (Tmax) of 200 °C, whereas that of the loaded sample was 180 °C. Furthermore, the portions of the curves for both the loaded and regenerated samples nearly coincide above 200 °C. This strongly suggests that the fraction of EG remaining on the surface after regeneration resides on higher-energy sites. Integration of the TGA desorption profiles shows that methanol regeneration removed 48% of the originally loaded EG. The percent removal was observed to increase with loading in subsequent experiments,4 in a way that suggests that the carbons have a relatively constant density of high-energy sites. After all of these highenergy sites are occupied, all subsequent solute mol-

Figure 4. TGA derivative plots for (a) wet MeOH-regenerated F400/HT carbon, (b) dry MeOH-regenerated F400/HT carbon, and (c) bare carbon reference. Original sample was EG-loaded (q ) 155 mg/g) F400/HT carbon. Heating rate: 10 °C/min.

ecules adsorbing onto the remaining low-energy sites should be amenable to solvent regeneration. “Wet” versus “Dry” Regeneration. Figure 4 compares the TGA desorption profiles for F400/HT carbon, loaded with EG (q ) 155 mg/g), with methanol as the solvent in wet and dry regeneration conditions. Several trends are immediately evident. First, it appears that the desorption profiles of the regenerated carbons are not substantially affected by the presence of water within the carbon. Also, there appears to be a larger amount of solute desorbing between 150 °C and 180 °C, which was not the case for the results for low loadings (q ) 23 mg/g) seen previously in Figure 3. This result is consistent with the notion that a larger proportion of the total solute occupies low-energy sites at high load-

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999 3749

Figure 5. TGA derivative plots for (a) loaded (from 0.82 wt. % EG) F400/HT carbon, (b) wet MeOH-regenerated F400/HT carbon, (c) reloaded (from 0.82 wt. % EG) F400/HT carbon, and (d) bare carbon reference. Heating rate: 15 °C/min.

ings. The fixed ratio of regenerant to carbon used (10 mL/g) was also probably insufficient for desorbing all of the glycol from the low-energy sites in the more heavily loaded carbon sample. Batch Regeneration with Reloading. To explore carbon regenerability and surface heterogeneity further, experiments were conducted in which the carbons were reloaded after being regenerated by wet methanol. Before reloading, the spent carbons were vacuum-dried to remove excess methanol. Figure 5 shows the TGA derivatives for each type of heat-treated F400 carbon: bare, loaded (from 0.82 wt. % EG), regenerated (by wet methanol), and reloaded (from 0.82 wt. % EG). As in Figure 3, a noticeable EG peak is present around 180 °C in the loaded and reloaded samples, followed by a high-temperature tail that extends up to 500 °C. Upon regeneration, most of the solute that originally desorbed at 180 °C has vanished. The EG remaining on the regenerated carbon remains on high-energy sites, which are manifested by the higher temperature range for desorption; that is, the tailing region in the derivative plots. When the carbon is reloaded, all of the additional solute appears exclusively in the regions near 180 °C, whereas the derivative values in the tailing regions do not change significantly. This confirms that nearly all of the high-energy sites were already occupied. Integrations of the desorption profiles for the loaded and reloaded carbon show similar values of EG uptake, which means that essentially the same equilibrium is reached from both stages. Process Implications. All of these findings suggest that after loading the carbons from aqueous solution, regeneration by solvents may be carried out in situ without the need for a preliminary drying step. As batch-wise contact of solvent appears to lead to incomplete removal, regeneration should be done with a large excess of solvent or using a continuous-contact device. The TGA results show that carbon may be reloaded to the same equilibrium capacity, provided that excess solvent is effectively removed upon regeneration.

Figure 6. Elution curves (25 °C) for solvent regeneration of beds packed with F400/HT carbon. Operating parameters: EGAcetone: Cfeed ) 0.0458 g/mL, F ) 1.58 mL/min, 1 BV ) 8.05 cm3. Propylene glycol (PG)-Acetone: Cfeed ) 0.0505 g/mL, F ) 1.54 mL/ min, 1 BV ) 6.68 cm3. Glycerol-Methanol: Cfeed ) 0.0275 g/mL, F ) 1.64 mL/min, 1 BV ) 6.21 cm3.

Fixed-Bed Regeneration: Focusing. Figure 6 shows representative elution curves for EG, PG, and glycerol during fixed-bed solvent regeneration of heattreated F400 carbon. The curves are of a peculiar shape. The first bed volume of fluid corresponds to the displacement of the aqueous feed solution (for which C/Cfeed ) 1) from the earlier breakthrough experiment. Afterward, the solute concentration begins to rise sharply to about three times the original aqueous-feed concentration. This effect is known as focusing,11 and is due to the large shift in adsorption equilibrium upon replacement of the original water with an organic solvent. The conditions in the presence of the solute now favor desorption, and the solute accumulates between the regenerant front entering and the aqueous front leaving, resulting in the large concentration peaks. Although not apparent from Figure 6, the peak for the solute occurs simultaneously with the breakthrough of the regenerant.4 When nearly all of the solute is finally removed from the bed, the effluent concentration falls to zero. Focusing Effect: Local Equilibrium Analysis. During solvent regeneration, the concentration profile of the solute can be predicted using local equilibrium analysis.11 It is assumed that, upon breakthrough of the regenerant, there is an abrupt change from a watersolute-carbon to a organic-solute-carbon adsorption equilibrium. From local equilibrium analysis, one may predict the new solute concentration due to the focusing effect, and the degree of tailing in the concentration profile after the focusing concentration is reached. As an example, we study the glycerol-methanol-F400/HT system shown in Figure 6. The parameters used for the analysis are listed in Table 1 below. The focusing concentration is determined from a mass balance over a differential length (∆z) of loaded bed, when in the time interval, ∆t, the solvent has changed from an aqueous to an organic one. The subscript 1 represents the (initial) conditions for the aqueous

3750

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999

Table 1. Local Equilibrium Analysis for Glycerol Recovery by Methanol Regeneration: Parameters

a

parameter

symbol

value

bed length (cm) interstitial porosity intraparticle porosity carbon skeletal density (g/cm3) interstitial velocity (cm/min) feed glycerol concentration (g/cm3) composite uptake of glycerol from water, q1 (g/g); C (g/cm3) composite uptake of glycerol from methanol, q2 (g/g); C (g/cm3)

L ee ep F v C1 q1 q2

17.2 0.46 0.57a 1.82a 4.54 0.0275 q ) 0.417C0.441 q ) 0.475C0.864

Typical value for coal-based carbons, taken from ref 11.

system, whereas 2 represents the (final) conditions for the organic system.

IN - OUT ) ACCUMULATION (IN SOLID + FLUID) (2) eev∆t(C2 - C1) ) ∆z[ee(C2 - C1) + (1 - ee)(1 - ep)F(q2 - q1)] (2a) The solvent velocity is the same as the interstitial velocity, hence ∆z/∆t ) v. After some simplification, we obtain:

(1 - ee)(1 - ep)F(q2 - q1) + ep(1 - ee)(C2 - C1) ) 0 (2b) The solute concentration at a point in the bed will be C1 until methanol passes through, when the concentration then increases to C2. The concentration increase will occur when sufficient time (L/v ) 3.7 min) has elapsed for methanol to pass through the entire bed. In the glycerol-methanol example, C2 was calculated to be 0.0748 g/cm3. Once the concentration C2 is reached, the adsorption equilibrium becomes unfavorable, as a solution of higher concentration (C2) is displaced from the bed by a solution of lower concentration (C ) 0). A diffuse wave results where the solute concentration will gradually decay to zero. From local equilibrium theory, the velocity (us) of the diffuse wave is concentration dependent, and has the form:11

us )

v 1 - ee 1 - ee dq2 1+ ep + (1 - ep)F ee ee dC

(3)

To determine the concentration profile, one calculates the time, t ) L/us, for an arbitrary range of concentrations between 0 and C2. Figure 7 compares the experimental elution profile with that predicted from local equilibrium theory for our glycerol-methanol-F400/HT example. There is quantitative agreement between C2 and the actual focusing concentration. However, as expected, there is substantial spreading observed at the front and rear of the concentration profiles. The profile shape at either end is likely attributable to axial dispersion, mass-transfer limitations, and the spreading of the solvent-water front. Because it is assumed from local equilibrium theory that the transition in adsorption equilibrium is instantaneous upon breakthrough of the solvent, the diffuse concentration profile calculated from eq 3 is much sharper than that observed from experiment. Optimization of Fixed Beds: Chromatography Theory. In practice, a sharper focusing effect is advan-

Figure 7. Local equilibrium analysis for fixed-bed methanol regeneration of glycerol from F400/HT carbon. Model parameters are listed in Table 1.

tageous for two reasons. A more concentrated product would lower downstream energy requirements for final purification substantially, and a lower number of bed volumes needed for full bed regeneration would reduce the amount of recycle of nonproduct solution required. Using some of the concepts of chromatography theory, we conducted a series of experiments designed to enhance the focusing effect through better performance of the fixed bed. In chromatography theory, the efficiency of a packed column is greater the lower the value of HETP (height of a theoretical plate). One mathematical form for HETP is:12

HETP ) Hp + Hd + Hs + Hm

(4)

Hp is the contribution from nonuniform paths for flow:

Hp ) 2λdp

(4a)

where dp is the particle diameter and λ is a constant close to unity that can be reduced with narrower particle-size distributions. Hd is the contribution from axial dispersion:

Hd )

2γDm v

(4b)

where Dm is the diffusivity of the solute in the mobile

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999 3751

Figure 8. Comparison of the focusing effects for acetoneregeneration of EG from heat-treated and acid-oxidized F400 carbon. Operating parameters: Heat treated: Cfeed ) 0.0459 g/mL; bed capacity** ) 0.175 g/g, F ) 1.58 mL/min, 1 BV ) 8.05 cm3; Ox 50/2*: Cfeed ) 0.0534 g/mL; bed capacity** ) 0.175 g/g, F ) 1.56 mL/min, 1 BV ) 7.60 cm3; Ox 70/2*: Cfeed ) 0.0534 g/mL; bed capacity**) 0.139 g/g, F ) 1.51 mL/min, 1 BV ) 6.59 cm3; Ox 70/9*: Cfeed ) 0.0483 g/mL; bed capacity** ) 0.126 g/g, F ) 1.38 mL/min, 1 BV ) 7.61 cm3. *Ox a/b denotes acid oxidation at a °C for b hours. **Defined as total EG desorbed from bed per dry mass of carbon.

phase, v is the flow velocity, and γ is a tortuosity factor (>1) related to the column packing. Hs and Hm are contributions from mass-transfer rates in the solid and fluid phase, respectively:

Hs )

QRD2v Ds

(4c)

ωdp2v Dm

(4d)

Hm )

where Q is a configuration factor that depends on the shape of the pool of stationary phase, R is a constant dependent upon the relative migration rate of solute in the mobile phase, D is the thickness of the stationary phase, Ds is the diffusivity of the solute in the stationary phase, and ω is the column coefficient that decreases when the bed packing is regular and tightly packed. Equation 4 indicates that a more efficient use of the bed could be achieved by finding an optimum volumetric flow rate and through the choice of carbon, by using smaller or more uniform particle sizes. Fixed-BedRegeneration: EffectofCarbonChoice. Organic solvents such as acetone or methanol adsorb to a much greater extent than do the multiple -OH solutes.4 Because of the capture of the solute in the front between the aqueous feed and the solvent displacer, the carbon that leads to the greatest degree of focusing during regeneration is likely to be the one with the highest equilibrium capacity; that is, a carbon of low oxygen content. Figure 8 shows acetone regeneration of EG-loaded beds packed with various types of F400 carbons.

Figure 9. Effect of volumetric flow rate on the focusing effect of EG from F400/HT carbon.

As expected, lower carbon capacity resulting from progressive acid oxidation leads to lower degrees of focusing. Another interesting feature of Figure 8 is that the Ox 50 °C/2 h bed seems to have a larger degree of tailing than the heat-treated bed. This suggests that in addition to lowering the equilibrium capacity, surface oxidation may cause a more sluggish water-to-acetone transition during focusing because of the greater affinity of the surface for water. Fixed-Bed Regeneration: Effect of Flow Rate. A series of heat-treated F400 beds (19 ( 2 cm) were loaded from a 5 wt. % EG solution, and subsequently regenerated by acetone at flow rates from 3.0 (0.33 mL/ min) to 34 (3.64 mL/min) BV/h. As seen in Figure 9, a focusing effect of 3 to 3.3-fold was observed for flow rates between 3 and 15 BV/h, whereas larger flow rates (2434 BV/h) resulted in a focusing effect of about two fold. In addition, at these higher flow rates, there was a greater degree of tailing. These results suggest that mass-transfer limitations are more controlling than axial dispersion. Sufficient residence time is obviously required for transport of the acetone into the carbon, and diffusion of the displaced solute from the carbon into the liquid stream. A similar result was observed previously,5 where for methanol regeneration of phenolloaded Pittsburgh CAL (carbon), a reduction of flow rate from 20 mL/min to 6 mL/min resulted in a two fold increase in focusing. In our work, 3-15 BV/h appears to be the likely optimum range of flow during regeneration, as lower flow rates may be discounted as tedious on the laboratory scale and likely impractical on the industrial scale. Fixed-Bed Regeneration: Effect of Particle Size/ Particle-Size Distribution. An experiment was conducted to compare the focusing effect for a bed packed with as-received F400/HT carbon (dp ) 0.45 to 2 mm) with that of another batch that had been sieved to a narrower particle size (dp ) 0.45 to 0.85 mm). Both beds were loaded with EG to similar degrees, and regenerated by acetone within the optimum range of flow. The results are shown in Figure 10, where a slight improvement of focusing (∼10% higher) was realized with the

3752

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999

Figure 10. Effect of particle-size distribution on the focusing effect of EG from F400/HT carbon.

Figure 11. Comparison of EG focusing by acetone for beds packed with Ambersorb 572 and F400/HT carbon. Operating parameters: F400/HT: bed capacity ) 0.205 g/g, Cfeed ) 0.0510 g/mL, F ) 0.67 mL/min, 1 BV ) 6.43 cm3; A572: bed capacity ) 0.231 g/g, Cfeed ) 0.0575 g/mL, F ) 1.11 mL/min, 1 BV ) 6.50 cm3.

more uniform particle sizes. In addition, and probably more important, the degree of tailing for the sieved particles was also reduced relative to the as-received batch. A sharper focusing effect might be obtained with an even narrower particle-size distribution. In a parallel experiment, acetone regeneration of EGloaded beds was carried out for two different adsorbents, F400/HT and Ambersorb 572. The differences in the EGfocusing effect between F400/HT and A572 shown in Figure 11 are striking. Although the equilibrium capacity of A572 for EG is no more than 50% higher than F400/HT,4 in a fixed-bed mode a focusing effect of six fold was observed for A572. Whereas the F400/HT had

Figure 12. Bed capacity of F400/HT carbon for EG up to 10 loading cycles, with acetone as regenerant.

a particle-size distribution between 0.45 and 2.0 mm diameter, the particles of A572 are all nearly spherical, with diameters between 0.25 and 0.45 mm. These results confirm that particle size and size uniformity substantially influence the focusing effect. Also, only 2.5 bed volumes are required for full removal of EG using A572 compared with 3 bed volumes with the best carbon. However, the price of the adsorbent is also important. Although A572 has a focusing effect twice as large as the best carbon, A572 is currently 10 to 20 times more expensive than conventional carbons on a per-pound basis. Fixed-Bed Regeneration: Cyclic Operation. Another important issue is whether the capacity and completeness of regeneration can be sustained through many cycles with the same bed. Two sets of experiments were performed on beds packed with F400/HT, with EG as the solute, using acetone and methanol as the respective regenerants. In both sets of experiments, the regenerated beds were flushed with 20 to 30 bed volumes of water per cycle to remove solvent in preparation for the next loading stage. Except for the choice of regenerant, all other operating parameters for the two beds were similar: bed length of about 19 cm, flow rates of 0.26 to 0.5 mL/min, and an aqueous feed solution containing 5 wt. % EG for all cycles. Figure 12 summarizes the results for 10 cycles of EG adsorption and acetone regeneration for a bed packed with F400/HT. The EG capacity was highest for the first cycle, but diminished by 30% on the subsequent cycles. Cycles 2 through 10 show a nearly constant bed capacity, with slightly lower values for the eighth and ninth cycles. The reason for the observed capacity losses is probably the difficulty in completely removing the acetone from the bed (refer to Figure 2). This acetone remains on the bed after the preceding wash cycle, which is insufficient to remove the acetone completely. During the loading cycles, the residual acetone from the bed was displaced by the adsorption of EG, resulting in a small focusing effect of acetone due to the shift in acetone-carbon to acetone-EG-water-carbon equilibrium.4 In a related study which 5 psig steam rather

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999 3753

appears to be superior to acetone as a regenerant because the latter remains on the carbon surfaces up to 200 °C in TGA measurements. Fixed-bed adsorption and regeneration demonstrated that glycols may be focused to peak concentrations three to six times the original feed concentration. This focusing effect can be optimized by using carbons of low oxygen content (e3 wt. %), low volumetric flow rate (315 BV/h), more nearly spherical particles, and more uniform particle sizes. Acknowledgment We thank Dr. James Kilduff, now Assistant Professor of Environmental Engineering at Rensselaer Polytechnic Institute, for sharing his data on adsorbent properties and for helpful discussions. Claire Law, an undergraduate in the Department of Chemical Engineering, University of California, Berkeley, assisted with experiments. This research was supported by the Assistant Secretary for Energy Efficiency and Renewable Energy, Office of Industrial Technologies, Division of Advanced Industrial Concepts of the U.S. Department of Energy under Contract No. DE-AC03-76SF00098. Figure 13. Bed capacity of F400/HT carbon for EG up to four loading cycles, with methanol as regenerant.

than room-temperature water was used to remove solvent, 15% capacity loss was observed after the first cycle for acetone regeneration of phenol-loaded Witco 940 carbon.8 Different results occur when methanol rather than acetone is chosen as the regenerant. As shown in Figure 13, the bed capacity is sustained through four cycles. Also, no residual methanol was detected in the effluent stream during any loading cycle, which suggests that methanol was fully removed during the water-wash stage. The ease of removing methanol relative to acetone is not surprising in view of Figure 2. Methanol was also regarded as the optimum solvent for cyclic fixed-bed recovery of phenol from Pittsburgh CAL (Calgon).5 However, because those authors did not use an excess of water during the rinse stage, a 12% capacity loss was observed after the first cycle. Conclusions Solvent regeneration was found to be an effective means of recovering adsorbed multiple -OH compounds from carbons. TGA, a precise tool for analyzing the adsorbed phase, confirmed that the completeness of regeneration was not affected by entrained water. Although a small fraction of solute appeared to be resistant to solvent regeneration, TGA showed that carbons may be regenerated to a large degree and subsequently reloaded without capacity loss. Methanol

Literature Cited (1) Chinn, D.; King, C. J. Adsorption of Glycols, Sugars, and Related Multiple -OH Compounds onto Activated Carbons. 1. Adsorption Mechanisms. Ind. Eng. Chem. Res. 1999, 38, 3738. (2) McLaughlin, H. S. Regenerate Activated Carbon Using Organic Solvents. Chem. Eng. Prog. July 1995, 45. (3) Himmelstein, K. J.; Fox, R. D.; Winter, T. H. Activated CarbonsIn-Place Regeneration of Activated Carbon. Chem. Eng. Prog. 1973, 69, 65. (4) Chinn, D.; King, C. J. Recovery of Glycols, Sugars, and Related Multiple -OH Compounds from Dilute-Aqueous Solution by Regenerable Adsorption onto Activated Carbons; Lawrence Berkeley National Laboratory Report No. 42904; Lawrence Berkeley National Laboratory: Berkeley, CA, 1999. (5) Cooney, D. O.; Nagerl, A.; Hines, A. L. Solvent Regeneration of Activated Carbon. Water Res. 1983, 17, 403. (6) Kilduff, J. E.; King, C. J. Effect of Carbon Adsorbent Surface Properties on the Uptake and Solvent Regeneration of Phenol. Ind. Eng. Chem. Res. 1997, 36, 1603. (7) Martin, R. J.; Ng, W. J. Chemical Regeneration of Exhausted Activated Carbon II. Water Res. 1985, 19, 1527. (8) Sutikno, T.; Himmelstein, K. J. Desorption of Phenol from Activated Carbon by Solvent Regeneration. Ind. Eng. Chem. Fundam. 1983, 22, 420. (9) Deyrup, C. Chemical Prices. Chem. Mark. Rep. 1998, 254. (10) Gmehling, J.; Onken, U.; Rarey-Nies, J. R. Vapor-Liquid Equilibrium Data Collection; DECHEMA: Frankfurt, 1988. (11) Wankat, P. C. Rate-Controlled Separations; Elsevier Applied Science: New York, 1990. (12) Johnson, E.; Stevenson, B. Basic Liquid Chromatography; Varian Associates: Palo Alto, 1978.

Received for review April 12, 1999 Revised manuscript received July 6, 1999 Accepted July 8, 1999 IE990289X