2416
Ind. Eng. Chem. Res. 1998, 37, 2416-2425
Effect of Nanopore Size Distributions on Trichloroethylene Adsorption and Desorption on Carbogenic Adsorbents Michael S. Kane, James H. Bushong, and Henry C. Foley* Center for Catalytic Science and Technology, Department of Chemical Engineering, University of Delaware, Newark, Delaware 19716
William H. Brendley, Jr. Department of Chemistry, Philadelphia College of Textiles, Philadelphia, Pennsylvania 19144
Two carbon adsorbents, Ambersorb-600 and Ambersorb-563 (A-600 and A-563), were compared for vapor-phase trichloroethylene (TCE) adsorption from humid air streams. These adsorbents retained capacity for TCE in humid environments and were regenerable in situ. Enhanced desorption, and hence, increased working capacities, were achieved with bimodal pore size distributions and hydrophobic surface chemistry. Vapor-phase TCE isotherms confirmed that both of these adsorbents have high capacities for TCE. Only a small difference between the micropore size distributions of A-563 and A-600 was determined by room-temperature methyl chloride adsorption and the modified Horvath-Kawazoe model. Besides differences in particle size and pore volume there was a measurable, but small change, in the fraction of the pores in the ultramicropore range (5 Å or smaller) of the A-600 adsorbent versus that of A-563. In packedbed breakthrough curve experiments, A-600 displayed a sharper mass-transfer zone than A-563, but maintained essentially the same capacity for TCE in a humid environment. Both materials were amenable to in-situ regeneration, and the A-600 a provided higher overall working capacity than that of A-563. Introduction Process waste stream cleanup often involves the removal of volatile and semivolatile organic solvents such as benzene, toluene, and halogenated organics such as trichloroethylene. To be successful, the pollutants must be reduced to acceptable levels in the parts per million (ppm) or in some cases the parts per billion (ppb) range.1 Adsorption processes are frequently used to remove and concentrate the organic for incineration or disposal.1-3 Because of its low cost and availability, activated carbon is the adsorbent most commonly used for vapor and liquid pollutants.4 Yet, activated carbon adsorption is not without problems, since the pollutantsaturated carbon becomes a hazardous material requiring special disposal procedures. If it is to be regenerated, this must be done ex-situ and most often off-site.5 Thus, despite the low cost of activated carbon, its continued use is subject to the increasing expense associated with its removal and disposal. It is these factors that fuel research aimed at new adsorbent materials designed for high working capacity, long life, and in-situ regenerability. Another important characteristic for any new adsorbent to have is the retention of capacity at high humidity.6 There is a competitive adsorption between water and the organic to be removed over activated carbon that results in the loss of capacity for the targeted organic.7 Thus, new adsorbent materials should be engineered to operate at high humidity if they are to offer further advantages over activated carbon. * To whom correspondence should be addressed. Phone: (302) 831-8056. Fax: (302) 831-2085. E-mail: foley@ che.udel.edu.
One approach to adsorbent material design is to prepare them not from natural precursors, but rather from polymers.8 Polymers, as opposed to natural precusors, offer the opportunity to more precisely control key properties, such as the micropore size distribution, the presence of transport pores, and surface hydrophobicity. However, although materials of this kind can be prepared in experimental quantities, particularly in academic laboratories, they generally have not been available on a commercial scale, with a few notable exceptions such as Carbosieve G and the Ambersorbs. While Carbosieve G is derived from pyrolysis of Saran, a blend of poly(vinylidene chloride) and poly(vinyl chloride), the Ambersorb adsorbents, are derived from the sulfonated styrene-divinylbenzene copolymer. These polymers are prepared by emulsion polymerization and used as ion-exchange resins. In addition to good mechanical and thermal stability and in-situ regenerability, these adsorbents can be made to have either net hydrophobic or hydrophilic surface chemistry9-11 along with controllable micro- and macropore size, providing a range of highly tunable properties.6 The results described in this work compare and contrast the behavior of two of the Ambersorb adsorbents, A-563 and A-600, for trichloroethylene (TCE) adsorption from a humid air stream. These findings are interpreted in light of the differences in their physical properties. Experimental Methods The Ambersorb 563 (A-563) and Ambersorb 600 (A600) adsorbent samples were provided by the Rohm and Haas Co. and were used without modification. Relevant physical properties for these two adsorbents are summarized in Table 1.
S0888-5885(97)00745-8 CCC: $15.00 © 1998 American Chemical Society Published on Web 05/07/1998
Ind. Eng. Chem. Res., Vol. 37, No. 6, 1998 2417 Table 1. Physical Property Data for A-563 and A-600 Adsorbents
adsorbent
particle size (mm)
bulk density (g/cm3)
mesoporosity (cm3/g)
macroporosity (cm3/g)
A-563 A-600
0.45 0.65
0.53 0.547
0.14 0.14
0.23 0.23
Pure-component, vapor-phase trichloroethylene adsorption and desorption isotherms were measured on a C.I. microbalance fitted with 0-10- and 0-760-Torr vacuum pressure gauges and a turbo-molecular pump backed by a two-stage rotary vane mechanical pump. This microbalance has a sample capacity of 100 mg and an ultimate sensitivity of (5 µg. Prior to the uptake measurements being taken, the samples were pretreated by heating them overnight at 300 °C in a dynamic vacuum of 1 × 10-6 Torr. Heat was provided to the sample by way of a copper tube furnace positioned around the sample hang-down tube and connected to a temperature controller. The trichloroethylene, TCE (Aldrich, 99.5+%, ACS reagent grade), was purified with multiple freeze-pump-thaw cycles to remove dissolved gases and to produce pure vapor for introduction into the microbalance. During the uptake experiments time, pressure, and mass were logged to an on-line computer. The TCE isotherm was obtained in the pressure range from 0.01 to 71 Torr at 25 °C. The maximum pressure corresponds to a TCE relative pressure of 0.92 (P0 ) 77 Torr at 25 °C). Photoacoustic infrared spectra were obtained with an MTEC model 200 photoacoustic detector installed in a Nicolet 510M FTIR spectrometer. The optical bench was purged with moisture-free and CO2-free air. The sample chamber of the photoacoustic detector was purged with high-purity helium (Keen Gas Co., 99.999%). A total of 500 co-added scans were collected for each sample. The adsorbent particles were placed directly in the spectrometer without grinding or preconditioning. Methyl chloride adsorption isotherms were obtained on the same C.I. microbalance used for the TCE adsorption experiments. Prior to these uptake measurements being taken, the samples were pretreated by heating them overnight at 300 °C in a dynamic vacuum of 1 × 10-6 Torr. Fresh adsorbent samples were used for each experiment; that is, the samples that were utilized for TCE uptake measurements were not reused for the methyl chloride adsorption studies. The methyl chloride adsorption isotherm was obtained in the pressure range from 1.3 to 700 Torr at 22 °C. The maximum pressure corresponds to a methyl chloride relative pressure of 0.18 (P0 ) 3904 Torr at 22 °C). Transformation of the methyl chloride isotherm into a pore size distribution was done by means of the Horvath-Kawazoe model.12 This model provides a relationship between the relative pressure and the micropore size, taken as the distance between the parallel walls in an idealized slit-shaped pore. In this way the methyl chloride isotherm data can be converted into an integral pore size distribution. The numerical derivative of this function provides the differential pore size distribution, which can be plotted as the number density of pores as a function of effective pore size from approximately 4.5 to 12 Å. The model is only appropriate to apply in this region of very small pores where the adsorbate diameter is not less than approximately half the pore size. Further details of this procedure can be found in a previous publication.13 Trichloroethylene breakthrough curves on each of the
adsorbents were determined with a bench-scale, fixedbed microadsorber unit (Figure 1). These adsorption breakthrough curves were measured for vapor-phase TCE adsorption from a humid air stream. The adsorbent bed consisted of approximately 75 g of either A-563 or A-600 held in place vertically with glass wool plugs and maintained at room temperature. For adsorption the flow conditions utilized were 10 L/min of total flow corresponding to an empty-bed superficial velocity of 19.7 ft/min through a bed that was 150 cm3 in total volume. The air flow, which was monitored with a calibrated rotometer, was split with a pair of metering valves with part of the flow sent to two gas spargers submerged in TCE and the bulk of the flow passing through a large water bubbler. All three bubblers were maintained at room temperature. The two streams were then recombined and the humidity of the water and TCE-containing stream was measured and maintained at greater than 95% relative humidity. A bypass line was provided to direct the feed stream around the bed before the experiment began in order to analyze and set the TCE concentration at the desired level without exposing the bed to TCE. A pressure gauge located upstream of the bed was used to measure the pressure drop through the bed for a pure air flow of 10 L/min. The metering valve on the bypass line was then adjusted to give the bypass line an identical pressure drop. At this point the metering valve leading to the TCE bubbler was opened to allow the inlet concentration to be set. This was done so that, at the start of the experiment, when the flow was switched from the bypass line to the adsorption bed, the inlet concentration would not change due to a change in the back pressure. The inlet TCE concentration was measured by withdrawing 10-µL gas samples manually through a septum placed upstream of the bed and injecting into a gas chromatograph equipped with a flame ionization detector (Gow Mac model 750P) and a 6 ft × 1/8 in. 100/120 Supeloport/ 10% SP-2100 column. Calibration standards for this gas chromatograph were obtained by injecting known quantities of pure liquid TCE into 20-mL gas sampling vials and allowing the TCE to completely vaporize. The flow of air to the TCE bubblers was adjusted until the desired 2000-ppmv inlet TCE concentration was obtained. At this point the adsorption experiment was started by switching the flow from the bypass line to the adsorption bed. Throughout the course of the adsorption experiment, the inlet and effluent concentrations of TCE were measured via 10-µL gas samples withdrawn from septa upstream and downstream of the bed, respectively. Samples were continually withdrawn until complete breakthrough was obtained (i.e., until the effluent and inlet concentrations were the same within experimental error). After the bed had been saturated completely, the adsorption experiment was stopped and the bed was regenerated thermally. For desorption, the bed was heated at 5 °C/min from room temperature to 180 °C and then held at 180 °C via a heating tape connected to a ramp and soak temperature controller. During the initial ramp up in temperature a vacuum pump was used to lower the pressure in the adsorber to approximately 2 in. of Hg. After the 30-min temperature ramp the vacuum pump was turned off and a nitrogen purge was flowed up through the bed at 2.5 L/min. As in the adsorption phase of the experiments, 10-µL gas samples of the effluent were withdrawn from a septum
2418 Ind. Eng. Chem. Res., Vol. 37, No. 6, 1998
Figure 1. Flow diagram for bench-scale adsorption unit. Table 2. Volume Fraction of Selected Pore Size Ranges in A-563 and A-600
Figure 2. Trichloroethylene vapor-phase adsorption and desorption isotherm on A-563 adsorbent. Temperature was 25 °C and TCE pressure ranged from 0.01 to 71 Torr.
located downstream from the bed and injected into the gas chromatograph for analysis. The effluent from the bed was scrubbed for acid, passed through a cold trap to remove organics, and then vented to a fume hood. Results TCE Vapor-Phase Adsorption and Desorption Isotherms on the A-563 Adsorbent. Gravimetric and vapor-phase adsorption and desorption isotherms for trichloroethylene on A-563 are shown in Figure 2. The isotherm displayed the pseudo-Type I Langmuir behavior characteristic of adsorption on a microporous material. Since the A-563 contains significant amounts of meso- and macroporosity (Tables 1 and 2), at higher relative pressures, capillary condensation and filling of the larger mesopores was observed. A-563 had a high
adsorbent
volume fraction 4-6 Å
volume fraction 6-8 Å
volume fraction >8 Å
A-563 A-600
53.0 49.7
26.5 25.3
20.5 25.0
capacity for TCE, even at fairly low relative pressures. At a relative pressure of 0.01 the capacity was roughly 200 mg/g; near saturation (P/P0 ) 0.92) the adsorption capacity was slightly over 400 mg/g. Using Gurvitch’s rule (density of TCE is 1.465 g/cm3), the pore volume occupied by TCE at a relative pressure of 0.92 is calculated to be 0.273 cm3. Methyl chloride adsorption indicates that A-563 has a micropore volume of 0.21 cm3/ g, and the mesopore volume is 0.14 cm3/g. Therefore, if all of the micropores are accessible to TCE, these adsorption results indicate that at a P/P0 of 0.92, all of the micropores and roughly half of the mesopores are filled with TCE. The desorption isotherm (Figure 2) shows that there is significant hysteresis at low relative pressures. The desorption isotherm begins to diverge from the adsorption isotherm at a relative pressure of 0.4. At progressively lower pressures, the difference between the adsorption and desorption isotherms increases. Even at extremely low relative pressures (P/P0 ) 1 × 10-4), the A-563 still retains 180 mg/g of TCE. Isotherm data indicate that A-563 has a high equilibrium capacity for trichloroethylene even at low relative pressures. Also of considerable practical importance is the rate of approach to equilibrium because the kinetics of adsorption and desorption frequently limit the performance of the adsorbent. The nanoporous nature of carbogenic materials often leads to significant intraparticle diffusion resistance. Figure 3 shows the rate of uptake of TCE on A-563, measured gravimetri-
Ind. Eng. Chem. Res., Vol. 37, No. 6, 1998 2419
Figure 3. Trichloroethylene uptake as a function of time on A-563 adsorbent. Temperature was 22 °C and TCE pressure was 0.01 Torr (P/P0 ) 1.3 × 10-4).
Figure 5. Trichloroethylene uptake on A-563 as a function of temperature during desorption. These uptake measurements were made under a dynamic vacuum.
Figure 4. Comparison of TCE isotherms measured from pure component vapor-phase adsorption and from aqueous liquid-phase adsorption. Vapor-phase relative pressures were converted into equivalent liquid-phase concentrations using the ideal gas law.
Figure 6. Plot of log of TCE uptake (in g of TCE/g of adsorbent) on A-563 versus inverse temperature (van’t Hoff equation). From the slope of this plot the TCE heat of adsorption can be calculated to be 4.3 kcal/mol.
cally, at a TCE pressure of 0.01 Torr. This figure shows that the TCE uptake is rapid, with equilibrium being achieved in roughly 1 h. Due to competitive adsorption, the presence of water can have a significant effect on the adsorbent’s capacity for the organic. Figure 4 compares the pure-component vapor-phase adsorption of TCE measured gravimetrically with TCE capacities measured from aqueous-phase adsorption. The liquid-phase concentrations were varied by adding known amounts of TCE to water and allowing the system to come to equilibrium. For comparison, the pure-component vapor-phase relative pressures were converted into equivalent liquid-phase concentrations using the ideal gas law:
concentration )
PTCE MWTCE RT
(1)
where PTCE is the vapor-phase partial pressure of TCE, R is the ideal gas constant, T is the temperature, and MWTCE is the molecular weight of TCE. These data show that at high concentrations the vapor- and liquidphase capacities were essentially the same. However, at lower concentrations (below 1 mg/L) the A-563 had a higher equilibrium adsorption capacity for TCE dissolved in the aqueous-phase than for the TCE from its own vapor phase.
Regeneration of the A-563 Adsorbent. The mass of TCE remaining on A-563 as a function of desorption temperature is shown in Figure 5. The first data point, taken at room temperature, was obtained after completion of the adsorption experiment by continuously pumping on the TCE-loaded sample. Even with continuous pumping at a dynamic vacuum pressure of 1 × 10-6 Torr, approximately 185 mg/g of TCE remained on the adsorbent after 45 min. The next three data points were obtained at higher temperatures and with continuous pumping. With each increase in temperature the desorption of TCE was fairly rapid. At 300 °C, the adsorbent was essentially completely regenerated with less than 5 mg/g of TCE remaining. The desorption data measured in this manner provides a measure of the pseudoadsorption equilibrium constants as a function of temperature. A van’t Hoff-like equation can be used to estimate the desorption energy by plotting the log of TCE mass (in g of TCE/g of adsorbent) on A-563 versus the inverse temperature (Figure 6). This plot is linear (R ) 0.993) with a slope of 2155 K-1. From this slope, the TCE heat of desorption can be calculated to be on the order of 4.3 kcal/mol (18 kJ/mol). The rate of desorption of TCE from A-563 as a function of time at room temperature is shown in Figure 7. Before this desorption experiment is begun, the
2420 Ind. Eng. Chem. Res., Vol. 37, No. 6, 1998
Figure 7. Uptake versus time for TCE desorption from A-563 at room temperature. Prior to the beginning of this desorption experiment, the A-563 adsorbent was equilibrated with 50 Torr of TCE (P/P0 ) 0.65).
Figure 8. Photoacoustic infrared spectra of fresh A-563 adsorbent and A-563 adsorbent containing adsorbed trichloroethylene.
A-563 adsorbent was equilibrated with 50 Torr of TCE. The A-563 has a capacity of 360 mg/g of TCE at this relative pressure (P/P0 ) 0.65). The desorption was carried out by continuous pumping on the sample. The TCE pressure dropped rapidly until a dynamic vacuum pressure of 1 × 10-6 Torr was achieved at room temperature. The sample was left in this condition for 48 h. This experiment shows that although the initial desorption was rapid, the TCE load dropped from 360 to 160 mg/g in the first hour. The desorption from the smallest pores was so slow that, even after 48 h in this dynamic vacuum, there remained 80 mg/g of TCE on the adsorbent, corresponding to 22% of the full capacity loading level. These data, coupled with the desorption data shown in Figure 5, indicate that temperatures of at least 150 °C are required to achieve full vacuum regeneration of the adsorbent. Photoacoustic Infrared Spectroscopy of Adsorbed TCE. Since a portion of the TCE was so slowly desorbed from the pore structure, the sample in that condition was removed from the microbalance and placed in a photoacoustic spectrometer to obtain an infrared spectrum of this strongly adsorbed species. Photoacoustic spectra of fresh A-563 and that of the sample containing the strongly adsorbed TCE are displayed in Figure 8. The peak at 3150 cm-1 is characteristic of an aromatic carbon-hydrogen stretch; the three bands observed at 880, 810, and 760 cm-1 are
Figure 9. Photoacoustic difference spectrum. Spectrum of A-563 adsorbent with adsorbed TCE minus spectrum of fresh A-563 adsorbent. Shaded areas represent the infrared absorption band locations for pure TCE.
the out of plane bending modes for the same hydrogens, and the broad band from 1700 to 1000 cm-1 is attributed to the carbon-carbon bonding in the backbone of the structure. The spectrum of A-563 with adsorbed TCE displays all of the same features plus additional infrared bands between 3000 and 2800 cm-1, 1000 and 700 cm-1, and at 1710 cm-1 that were not observed in the clean material. The differences between these materials can be more easily seen by looking at the photoacoustic difference spectrum. The difference of the spectrum for A-563 with adsorbed TCE minus the spectrum for fresh A-563 is shown in Figure 9. Also shown in this figure are the infrared absorption band locations for pure TCE, which are represented by the shaded gray bars on the figure. The first thing to note are that the three bands observed at 934, 840, and 721 cm-1 for TCE strongly adsorbed on A-563 coincide with the three most intense bands observed for pure TCE, but the other three absorption bands for TCE, at 3125, 1587, and 1250 cm-1, are not observed per se. Two of these bands (∼1750 and 1450 cm-1) fall within the broad carbon-carbon absorption region and could be shifted by the strong absorption of the carbogenic material in the pore. The third band for TCE (2900 cm-1) is roughly half as intense as the three peaks between 1000 and 700 cm-1 and also may be shifted by the same strongly absorbing force of the carbon pore. Methyl Chloride Adsorption and Pore Size Distribution Analysis of A-563 and A-600 Adsorbents. Methyl chloride adsorption isotherms for the A-563 and A-600 adsorbents (Figure 10) were quite similar in shape and nearly parallel. At the highest relative pressures investigated (P/Psat ) 0.18) the equilibrium adsorption capacities for methyl chloride were 190 mg/g for A-563 and 170 mg/g for A-600. Using Gurvitch’s rule (methyl chloride density is 0.92 g/cm3), the total micropore volumes for A-563 and A-600 are calculated to be 0.21 and 0.19 cm3/g, respectively. More importantly, by using the modified HorvathKawazoe model,13 this adsorption data can be used to calculate differential pore size distributions for these two materials. The distributions of micropore sizes in the region from 4.5 to 12 Å for these samples are shown in Figure 11. As we have reported before with other carbogenic adsorbents, the mode of the pore size distri-
Ind. Eng. Chem. Res., Vol. 37, No. 6, 1998 2421
Figure 10. Methyl chloride adsorption isotherms for A-563 and A-600 adsorbents plotted as a function of relative pressure (P/P0). Temperature was 22 °C and methyl chloride pressure ranged from 1.3 to 700 Torr.
Figure 12. (a) Trichloroethylene concentration breakthrough profiles as a function of time on stream for A-563 and A-600 adsorbents. Temperature ) 22 °C, air flow ) 10 L/min, inlet TCE concentration ) 2000 ppmv, greater than 95% relative humidity. (b) Trichloroethylene concentration breakthrough profiles for A-563 and A-600 adsorbents with time axis normalized by adsorbent bed mass. Figure 11. Differential pore size distributions for A-563 and A-600 adsorbents using the Horvath-Kawazoe model and methyl chloride as the probe molecule. Distribution plotted as population of pores as a function of effective pore size.
bution for both materials is approximately 5 Å. The average pore sizes between 4.5 and 12 Å for A-563 and A-600 are 6.55 and 6.76 Å, respectively (i.e., they are nearly identical). Although there are not major differences between these two distributions, there are some differences. The distribution of A-563 is more sharply peaked in the region between 4.5 and 6 Å, and the population of pores between 6 and 8 Å is slightly lower for A-600 than for A-563. Above 8 Å the pore size distribution of A-600 displays a slightly larger population of pores than does that of A-563. Overall, the A-600 shows a somewhat flatter pore size distribution with slightly lower pore volume for methyl chloride than A-563. TCE Breakthrough Profiles. The total capacities for trichloroethylene on A-563 and A-600 were measured at greater than 95% relative humidity and at 2000 ppmv TCE concentration from the adsorption breakthrough profile experiments. At these conditions, A-563 had a total uptake of 281 mg of TCE/g of adsorbent, and A-600 had a slightly lower total capacity of 267 mg of TCE/g of adsorbent. The TCE concentration breakthrough profiles corresponding to the two adsorbents operating under identical conditions are shown in Figure 12a. For the A-563 adsorbent, initial breakthrough of TCE, at approximately 10 ppmv, occurred after 130 min on
stream followed by a steady linear increase in effluent TCE concentration with longer times on stream. Complete breakthrough, defined as the point where the inlet and outlet concentrations are identical, was achieved after 4 h on stream. For the A-600 adsorbent initial breakthrough of TCE occurred after 175 min of time on stream followed by a sharper increase in effluent TCE concentration with increasing time on stream, as compared to that of A-563. Complete breakthrough on this material also was achieved after 4 h on stream. Since the masses of the two beds were not the same, the time axis of the breakthrough curves has been normalized by bed mass so that on this basis a more valid comparison of adsorption behavior can be made. These normalized breakthrough profiles are shown in Figure 12b. The data show that the breakthrough curve for A-600 was sharper than that for A-563; initial breakthrough occurred later for the A-600 with the concentration profiles crossing over at an effluent TCE concentration of approximately 500 ppmv. In-Situ Regeneration. The adsorbents were regenerated via thermal desorption by heating the bed at 5 °C/min from room temperature to 180 °C. The effluent TCE concentration profiles during this thermal desorption are shown in Figure 13. For both adsorbents, the desorption profile was fairly sharp with the greater portion of TCE being desorbed during the first 50 min. The effluent concentration for both materials rose rapidly as soon as vacuum and heating were applied.
2422 Ind. Eng. Chem. Res., Vol. 37, No. 6, 1998 Table 3. Depth and Speed of Mass-Transfer Zones in TCE Concentration Breakthrough Profile Experiments on A-563 and A-600 Adsorbents adsorbent
S (cm/min)
DMTZ (cm)
A-563 A-600
4.3 × 10-2 4.4 × 10-2
4.8 2.9
Table 4. Physical Characteristics and Length of Adsorption Beds for Breakthrough Profile Experiments on A-563 and A-600 Adsorbents
Figure 13. Trichloroethylene desorption concentration profiles for A-563 and A-600 adsorbents. Bed heated at 5 °C/min to 180 °C.
In both cases, the effluent concentration dropped and then spiked upward when the nitrogen purge was started after 30 min of desorption. The effluent concentration peaked at approximately 75 000 ppmv for A-600 and at 30 000 ppmv for A-563. This higher peak for A-600 could be attributed to the fact that the A-600 bed mass was 14 g larger than that for A-563. The effluent TCE concentrations for both adsorbents fell below 500 ppmv within 90 min of desorption. Even though the particle size of the A-600 is nearly 50% larger, both adsorbents display similar desorption profiles and are largely regenerated within the first 90 min of desorption. Analysis Mass-Transfer Zone Calculation. The TCE concentration breakthrough profiles demonstrate the effect of the shift in porosity on the adsorption behavior of these adsorbents (Figure 12a,b). Because of this, even though the total saturation capacity at 2000 ppmv TCE concentration was lower for A-600, it took longer for initial breakthrough to occur. On A-600 the breakthrough curve was much sharper than that for A-563, indicating a sharper mass-transfer zone and hence more effective utilization of the adsorbent across the bed. The breakthrough profiles can be analyzed in more detail by calculating the depth and speed of propagation of this adsorption mass-transfer zone in the bed. The speed of propagation6 of the mass-transfer zone is defined as
S)
C0F LsF0A
(2)
where S is the speed of the mass-transfer zone (cm/min), C0 is the inlet concentration of TCE (mg/L), F is the total flow rate (L/min), Ls is the saturation capacity of the adsorbent (mg of TCE/g of adsorbent), F0 is the bulk density of the adsorbent (mg/g), and A is the crosssectional area of the adsorption bed (cm2). From this, the depth of the mass-transfer zone in the bed is calculated as
DMTZ ) S(ts - tb)
(3)
where DMTZ is the depth of the mass-transfer zone (cm),
adsorbent
bed mass (g)
density (g/cm3)
bed area (cm2)
length of bed (cm)
A-563 A-600
69.6 83.5
0.53 0.547
16.7 16.7
7.9 9.2
ts is the time required for complete saturation of the bed (min), and tb is the time at initial breakthrough (min). The results of these calculations are shown in Table 3. The speed of propagation of the mass-transfer zone was essentially the same for the A-563 and the A-600 in these experiments. However, the depth of the masstransfer zone for A-600 was only 2.9 cm, nearly a factor of 2 smaller than the 4.8-cm zone calculated for A-563. This much smaller mass-transfer zone was evidence of a more favorable mass-transfer profile on A-600 that allowed for more effective utilization of the bed, even though the particle size for the A-600 was significantly larger. LDFQ (Linear Driving Force) Modeling of Breakthrough Profiles. The length of the adsorption beds in these experiments can be calculated from the crosssectional area of flow and the mass and density of the adsorbents. Relevant physical properties of the adsorbent and adsorption bed for A-563 and A-600 are given in Table 4. For A-563, the adsorbent bed was roughly 1.7 times longer than the depth of the mass-transfer zone. For A-600, which had a much sharper masstransfer zone, the adsorbent bed was slightly more than 3 times longer than the depth of the mass-transfer zone. The favorable (nonlinear) TCE adsorption isotherms observed on these adsorbents gave rise to the formation of constant pattern mass-transfer profiles in these breakthrough experiments. That is, the spreading tendency due to nonequilibrium adsorption or dispersion was counterbalanced by the sharpening tendency due to the nonlinearity of the favorable adsorption isotherm.14 For these adsorption experiments, a constant pattern mass-transfer profile can be assumed because the length of the adsorption bed was larger than the depth of the mass-transfer zone. These constant-pattern breakthrough profiles can be modeled with a linear driving force.6 The LDFQ model (linear driving force model based on particle-phase concentration differences) is applicable when intraparticle diffusion plays a dominant role in mass transfer. Because of the very high flow rates that were used in this system (10 L/min), external mass-transfer resistance from the fluid to the particle was not significant. (Calculation of the external mass-transfer coefficient using mass-transfer correlations for flow through a fixed bed of pellets verified that the external mass-transfer coefficient was at least an order of magnitude higher than the highest intraparticle-mass transfer coefficient.) The governing differential equation for mass transfer in the LDFQ model is given by
Ind. Eng. Chem. Res., Vol. 37, No. 6, 1998 2423
Fb
dq ) Ksav(q* - q) dt
(4)
where Fb is the bulk density of the bed (g/cm3), q is the particle-phase concentration of the adsorbate (mg/g), t is time (s), q* is the particle-phase concentration at equilibrium with the fluid-phase concentration (mg/g), and Ksav is the overall mass-transfer coefficient (mol/ cm3/s). When intraparticle diffusion is the dominant mass-transfer resistance, Ksav can be replaced by ksav, which is related to the intraparticle pore diffusion parameters as follows:
ksav )
15φsDp/R2 q0/C0
(5)
where Dp is the intraparticle diffusivity (cm2/s), R is the particle radius (cm), q0 is the particle-phase concentration in equilibrium with the inlet fluid-phase concentration (mg/g), C0 is the inlet fluid-phase concentration (mol/L), and φs is a parameter that accounts for the deviation from linearity of the adsorption isotherm and is a function of the isotherm parameters.14 For this study, the Langmuir and Freundlich isotherms were used in the LDFQ model. The Langmuir adsorption isotherm is given by
y)
x r + (1 - r)x
(6)
where y is the dimensionless particle-phase concentration, x is the dimensionless fluid-phase concentration, and r is the Langmuir isotherm parameter that measures the degree of nonlinearity of the isotherm (when r is unity, the Langmuir isotherm is linear). For this form of the adsorption isotherm, φs can be correlated with eq 8.15
φs ) 1 - 0.192(1 - r)3
(7)
The Freundlich adsorption isotherm is given by
y ) x1/n
(8)
where n is the Freundlich isotherm parameter that measures the degree of nonlinearity of the isotherm (when n is unity, the Freundlich isotherm is linear). For this isotherm equation, φs can be correlated using eq 5.15
φs ) 0.808 +
0.192 n
(9)
The governing differential equation for mass transfer (eq 4) can be solved explicitly in terms of the masstransfer coefficients, dimensionless concentrations, and isotherm parameters for both the Langmuir and Freundlich isotherms. For the Langmuir isotherm, the solution is16
ksav 1 r ln x ln(1 - x) - 1 (t - t0) ) Fb 1-r 1-r
(10)
and for the Freundlich isotherm, the solution is
ksav n [ln(1 - x(n-1)/n) + IB] (t - t0) ) Fb 1-n
(11)
Figure 14. Comparison of experimental trichloroethylene concentration breakthrough profiles and LDFQ (linear driving force) model. LDFQ model fits are shown for the Langmuir adsorption isotherm and the Freundlich adsorption isotherm.
where
∫01ln(1 - x(1-1/n)) dx
IB ) -
(12)
which can be approximated using the first 20 terms of an infinite series. The LDFQ model was applied to the experimental data by fitting eq 10 for the Langmuir isotherm and eq 11 for the Freundlich isotherm. For the Langmuir isotherm the two adjustable parameters were the intraparticle diffusivity, Dp, and the isotherm parameter, r. For the Freundlich isotherm, the adjustable parameters were Dp and the isotherm parameter, n. For each isotherm, optimum values of the adjustable parameters were sought using a Newton search method to minimize the sum of the least-squaresd errors between the experimental data and the model. Comparison of the experimental data and model are shown in Figure 14. Agreement between the model and the experimental data for both A-563 and A-600 is quite good, particularly for the Freundlich isotherm. The optimum values for the LDFQ model parameters are shown in Table 5. The optimum values for the isotherm parameters, r and n, are nearly identical for A-563 and A-600. This suggests that the shapes of the TCE adsorption isotherms are similar for these adsorbents, a reasonable conclusion based on a comparison of the methyl chloride adsorption isotherms on these materials. The fact that the regressed values for r are less than unity, and the values for n are greater than unity, confirms the favorable nonlinearity of the trichloroethylene adsorption isotherms. The effective intraparticle diffusivities for TCE in A-563 and A-600, determined from the LDFQ model, are on the order of 10-3-10-2 cm2/s. These high intraparticle diffusivities can be attributed to the significant amount of meso- and macroporosity in these adsorbents. For both A-563 and A-600, the effective diffusivities calculated for the Freundlich isotherm are approximately twice those determined from the Langmuir model. However, for both isotherm models, the ratio of the effective diffusivity in A-600 to the effective diffusivity in A-563 is between 3 and 4. This 3-fold larger effective diffusivity is further evidence for improved internal mass transfer in A-600 arising from small changes in the nanopore distribution compared to that of A-563. The net effect of this increased
2424 Ind. Eng. Chem. Res., Vol. 37, No. 6, 1998 Table 5. LDFQ Model Parameters. Langmuir and Freundlich Isotherm Parameters and Effective Intraparticle Diffusivities Determined by Minimization of Least Squares. Ratio of Effective Diffusivity for A-600 to Effective Diffusivity for A-563 Langmuir isotherm adsorbent A-563 A-600
r 0.22 0.23 ratio of diffusivities )
Freundlich isotherm Dp (cm2/s) 10-3
1.57 × 4.7 × 10-3 3.0
nanopore diffusivity, coupled with the Knudsen and molecular diffusion processes occurring both in parallel and in series in the larger pores, is to increase the total effective diffusivity by a factor of 3. Discussion The transport porosity and hydrophobic surface chemistry of these adsorbents are critical properties that yield improved performance for vapor-phase TCE adsorption. Figure 3 shows that TCE adsorption on A-563 reaches equilibrium in less than 1 h. This rapid uptake demonstrates the importance of the meso- and macroporosity on transport. Comparison of pure TCE vapor-phase adsorption with TCE adsorption from the aqueous phase (Figure 4) shows that there is higher capacity for TCE from the aqueous-phase in the low-concentration region. This suggests that the hydrophobic surface chemistry of the A-563 adsorbent plays a significant role in the adsorption of TCE in the presence of water.9 At higher concentrations of TCE, it is interesting to note that the extents of adsorption of TCE are nearly the same whether the TCE emerges from liquid water or from the pure vapor. The range of nonidealities that can be observed in vapor-phase coadsorption of organic compounds and water on activated carbon has been discussed by LeVan and co-workers.17 On the basis of their results, the observations made here with TCE and water coadsorption are less unexpected than they might otherwise have been. Both sets of results point to the intrinsic role of the adsorbent surface chemistry in the nature of the nucleation of the coadsorbed phase. Static vapor-phase adsorption of trichloroethylene on A-563 showed that this adsorbent not only has high capacity for TCE but also has low internal mass-transfer resistance. The desorption studies demonstrated the potential for in-situ regeneration at moderate temperatures. The photoacoustic spectroscopy of the more strongly adsorbed trichloroethylene indicates that chemical decomposition of TCE during desorption can take place in the pores. The methyl chloride isotherms indicate that the total microporosity is lower within A-600 compared to that within A-563 (Figure 10). Comparison of the two isotherms shows that most of the increased methyl chloride capacity on A-563 occurs in the low-pressure region (P/P0 < 0.01). Above this pressure level the isotherms are essentially parallel. An analysis of the pore size distribution of these two materials shows that the total microporosity of A-600 was slightly lower in the 4.5-7-Å region compared to that of A-563 (Figure 11). Even though the total porosity is less for A-600, there are relatively more micropores in the 7-12-Å region. Thus, for A-600 the smallest pores with the strongest adsorption energies were diminished in number. Despite nearly identical average pore sizes for A-600 and A-563, the volume fractions of pores in different size ranges were different (Table 2); for A-563 approximately 20% of the pores are greater than 8 Å in
n
Dp (cm2/s)
1.4 1.3 ratio of diffusivities )
3.7 × 10-3 1.4 × 10-2 3.8
size, while in A-600 this percentage is increased to 25%. When the pore size and molecule size are close in magnitude, the diffusion is configurational in nature and very size dependent. Therefore, the effective diffusivity within an 8-Å pore can be expected to be much larger than that in a 6-Å pore. Consequently, this small, but measurable, redistribution of pores to larger sizes may have a substantial impact on the overall effective diffusivity in the adsorbent particle. Keeping in mind that the meso- and macropore volumes were essentially the same as were the surface chemistries, then the only distinctions between these two materials were the small, but measurable, difference in ther micropore size distributions and particle size, which is 50% larger for A-600 than for A-563. The need for operating regeneration at mild conditions was made evident by the reactivity of the adsorbent observed in the photoacoustic difference spectrum presented in Figure 9. Here, the comparison is made between the photoacoustic spectrum of A-563 plus adsorbed TCE to the spectrum for fresh A-563. The most interesting infrared bands are those observed for TCE adsorbed on A-563 that cannot be attributed to pure TCE. Absorption bands at 1710 cm-1 and three peaks at 2950, 2925, and 2860 cm-1 were observed in the A-563 plus TCE sample, but not in the spectra of either the pure TCE or fresh A-563 sample. The band at 1710 cm-1 could be attributed to absorption by a carbonyl species, and the three peaks between 3000 and 2800 cm-1 are in good agreement with the symmetric and asymmetric carbon-hydrogen stretches observed for aliphatic carbons. However, there are no aliphatic carbons or oxygen in trichloroethylene and no measurable (via infrared spectroscopy) aliphatic carbons in the A-563 adsorbent. Although there is some oxygen present in the A-563 adsorbent in the form of sulfone or sulfonic acid groups,9 these would have to react with trichloroethylene to form a carbonyl-containing species. At room temperature, the forces exerted on the adsorbed trichloroethylene and the reactivity of the residual sulfone and sulfonic acid groups in these nanopores may lead to chemical rearrangement or even cracking of the trichloroethylene molecule, but this is not proven. Thus, it is more likely that the bands of pure TCE are split and shifted by the perturbing forces of the carbon nanopore. Hence, at the temperatures required for effective desorption, chemical decomposition of trichloroethylene with concomitant HCl formation could be a significant problem in the early stages of this adsorbent’s use. The desorption isotherms presented in Figure 2 showed that even at a low relative pressure of 1 × 10-4, the A-563 retained 180 mg/g of strongly adsorbed TCE. This behavior can be attributed to the strong energy of adsorption for TCE in the smallest pores and the resultant slow rate of diffusion out of these pores. The result also demonstrated the need for an increase in temperature to effectively desorb and regenerate the adsorbent. From Figure 6, the apparent heat of de-
Ind. Eng. Chem. Res., Vol. 37, No. 6, 1998 2425
sorption for TCE was calculated to be on the order of 4.3 kcal/mol, that is 18 kJ/mol. This is essentially the same as to the “mean adsorption potential” of 18.8 kJ/ mol for TCE on A-563 reported by Parker11 on the basis of adsorption rather than desorption experiments. The mass-transfer analysis showed the effect of shifting the nanopore size distribution on the adsorption behavior of these carbogenic materials. During desorption, the effect of the shift in nanoporosity in A-600 was also evident, and potentially much more important. The desorption from A-600 occurred over roughly the same time frame as that of the A-563 adsorbent. Even though the A-600 particle size is nearly 50% larger, both materials are nearly fully regenerated after 90 min of desorption. The characteristic time for diffusion is proportional to the square of the particle diameter. If the diffusivity remained constant, a particle that is 1.45 times larger could require on the order of twice as much time to be regenerated (1.452 ) 2.1). However, the desorption profiles on both of these adsorbents were quite similar which suggests that the effective diffusivity, and the rate of desorption, have been increased in the A-600, a conclusion that is in conformity with the results from the adsorption breakthrough studies. The net effect of this is that nearly equivalent in-situ regeneration cycles could be used for both adsorbents. This is particularly important in light of the photoacoustic infrared data which indicate the potential for TCE decomposition at elevated temperatures. With the improved desorption profile of A-600 it could be regenerated effectively without resorting to higher desorption temperature, minimizing the TCE decomposition. Conclusions With its larger particle size, A-600 can provide a lower pressure drop through the bed than A-563. Without any other change this change alone would have the effect of lengthening the time required to fully utilize the adsorbent. However, by reengineering the pore structure to favor the slightly larger, lower energy micropore sizes, the intraparticle resistance can be lowered. Thus, the adsorption mass-transfer zones are sharper and initial breakthrough occurs later, even though the overall saturation capacity is somewhat lower. Hence, more effective utilization of the adsorption bed can be made, especially if the bed is not allowed to reach full saturation, but rather the adsorption cycle is terminated when the effluent concentration reaches a specified level. For example, if a maximum effluent concentration of 10 ppmv of TCE were specified with an inlet TCE concentration of 2000 ppmv, then a 100-g bed of A-563 would be able to treat 1900 L before regeneration, whereas a 100-g bed of A-600 would be able to treat 2100 L. This 10% increase in working capacity comes in addition to a more favorable desorption profile; the improved desorption allows for shorter cycles times. Another finding of importance is that the capacities for TCE at greater than 95% relative humidity are high on both adsorbents, 281 mg/g for A-563 and 267 mg/g for A-600. These adsorbents are hydrophobic and actually display slightly higher capacities for TCE adsorption from a humid air stream than from pure TCE vapor. This could provide a working adsorber unit with higher efficiency at shorter cycle times. However, the beneficial properties of these synthetic, polymerderived adsorbents come at a much higher cost than that for activated carbon. Thus, when considering the
design of an adsorber, the much higher initial investment can only be justified if it is offset in time by much lower overall operating costs. At the same time it is important to recognize that higher working capacity translates into less adsorbent mass for the same level of performance and this could mitigate, somewhat, against the otherwise larger capital investment. If increasingly stringent environmental regulations could be factored into the economics, then the internal rate of return for this adsorbent technology might become acceptable in the future. Economic issues not withstanding, the new technology of nanoscale engineered adsorbents is real and it could have a significant impact on environmental processing in the not-too-distant future. Literature Cited (1) Environmental Science and EngineeringsRemoval of Volatile Organic Chemicals from Potable Water; Noyes Data Corporation: New Jersey, 1986. (2) Neely, J. In Activated Carbon Adsorption from the Aqueous Phase, Vol. 2; McGuire, M., Suffet, I., Eds.; Ann Arbor Science: Ann Arbor, MI, 1980; p 417. (3) Heilshorn, E. Removing VOCs from Contaminated Water. Chem. Eng. 1991, 98, 120. (4) Bansal, R.; Donnet, J.; Stoeckli, F. Active Carbon; Marcel Dekker: New York, 1988. (5) Cheremisinoff, N.; Cheremisinoff, P. Carbon Adsorption for Pollution Control; Prentice Hall: Englewood Cliffs, NJ, 1993. (6) Neely, J.; Isacoff, E. Carbonaceous Adsorbents for the Treatment of Ground and Surface Waters; Marcel Dekker: New York, 1982. (7) Cookson, T., Jr. In Carbon Adsorption Handbook; Cheremisinoff, P., Ellerbusch, F., Eds.; Ann Arbor Science: Ann Arbor, MI, 1978; p 241. (8) Foley, H. Carbogenic Molecular Sieves: Synthesis, Properties and Application. Microporous Mater. 1995, 4, 407. (9) Neely, J. Characterization of Polymer Carbons Derived from Porous Sulfonated Polystyrene. Carbon 1981, 19, 27. (10) Parker, G., Jr. Optimum Isotherm Equation and Thermodynamic Interpretation for Aqueous 1,1,2-Trichloroethylene Adsorption Isotherms. Adsorption 1995, 1, 113. (11) Parker, G., Jr. American Institute of Chemical Engineers Annual Conference, Miami, Florida, 1992. (12) Horvath, G.; Kawazoe, K. Method for the Calculation of Effective Pore Size Distribution in Molecular Sieve Carbon. J. Chem. Eng. Jpn. 1983, 16, 470. (13) Mariwala, R.; Foley, H. Calculation of Micropore Sizes in Carbogenic Materials from the Methyl Chloride Adsorption Isotherm. Ind. Eng. Chem. Res. 1994, 33, 2314. (14) Suzuki, M. Adsorption Engineering; Elsevier: New York, 1990. (15) Miura, K.; Hashimoto, K. Analytical Solutions for the Breakthrough Curves of Fixed-Bed Adsorbers Under Constant Pattern and Linear Driving Force Approximations. J. Chem. Eng. Jpn. 1977, 10, 490. (16) Hall, K.; Eagleton, L.; Acrivos, A.; Vermeulen, T. Pore- and Solid-Diffusion Kinetics in FixedBed Adsorption Under Constant Pattern Conditions. Ind. Eng. Chem. Fundam. 1966, 5, 212. (17) Taqvi, S. M.; LeVan, M. D. Nonidealities in Vapor-Phase Coadsorption of Organic Compounds and Water on Activated Carbon. In Fundamentals of Adsorption; LeVan, M. D., Ed.; Kluwer: Boston, 1996; p 971.
Received for review October 27, 1997 Revised manuscript received March 25, 1998 Accepted March 26, 1998 IE970745M