Environ. Sci. Technol. 1988,22, 313-321
Modeling Activated Carbon Adsorption of Target Organic Compounds from Leachate-Contaminated Groundwaters Edward H. Smith and Walter J. Weber, Jr.”
Environmental Engineering Program, The University of Michigan, Ann Arbor, Michigan 48 109 A relatively straightforward mathematical modeling technique employing a modified homogeneous surface diffusion version of the Michigan Adsorption Design and Applications Model was used to simulate and predict fixed-bed adsorber behavior with respect to two target organic compounds in a complex background water contaminated by leachate from a hazardous waste landfill. The approach applied was one in which model coefficients specific to the leachate mixture were evaluated for the target compounds and the leachate itself was considered only as unspecified background. The unspecified dissolved organic matter in the leachate reduced granular activated carbon adsorption capacities and rates for both target compounds relative to their respective values in waters containing no other organic species. A short-bed adsorber technique was employed for estimation of system-specific rate parameters for both single-componentand dual-component combinations of the target compounds with the leachate background. Introduction
Granular activated carbon (GAC) treatment is one of the fundamental technologies for prevention and remediation of groundwater contamination. Site-to-site variabilities and the complex composition of many contaminated groundwaters complicate the design and operation of adsorption systems for such applications, particularly in cases involving leachates from landfills and dump sites. In addition to specific “target” organic contaminants, leachates typically contain relatively high concentrations of uncharacterized organics, referred to generically as dissolved organic matter (DOM). Although treatment objectives may be directed toward one or two specific target compounds, the interactions and transformations occurring among these compounds and other organic species in solution must be properly acknowledged in the specification and engineering of appropriate adsorption systems. Increased contaminant solubility, decreased partitioning to adsorbing solids, unequal competition among orgafiic adsorbates, hysteresis, steric effects, and hindered diffusion are examples of interactions that have been observed, but reports are generally limited to singular and well-defined pollutant/humic material combinations (1-6). Pilot-scale testing traditionally has been utilized to obtain design information for activated carbon systems. More recently, mathematical modeling approaches that attempt to characterize a given physical system in a simplified but meaningful manner have been employed to reduce the number of pilot-scale testa required to evaluate variable operating conditions and ranges of design parameters. As indicated in a parent paper by the authors (7), substantial progress has been made in this regard over the past several years, and models now exist that can accurately predict interactions between trace contaminants and activated carbon and estimate breakthrough patterns for adsorbing solutes in effluents from fixed-bed adsorbers as a function of time or volume of water treated. There is, moreover, evidence that these models, if properly modified, offer potential as design tools for tailoring ad0013-936X/88/0922-0313$01.50/0
sorption processes to waters containing complex organic mixtures (1,2,6,8-10). This study focuses on the modification and application of a specific modeling methodology to such systems. Objectives and Approach
The several objectives of the research were to (1)examine the effects of waste leachate DOM backgrounds on adsorption capacities and rates for typical organic compounds, (2) apply an existing modeling technique for description and prediction of fixed-bed adsorber breakthrough characteristics for these target compounds in the presence of background DOM, and (3) evaluate several approaches for estimating adsorption rate parameters for the target compounds in the systems studied. A homogeneous surface diffusion (HSD) version of the dual-resistance mass-transfer Michigan Adsorption Design and Applications Model (MADAM) was used to simulate and predict fixed-bed adsorber performance (11, 12). Required input to the MADAM-HSD model includes, for each compound modeled, a film diffusion parameter kf relating to the mass transfer of solute to the exterior carbon particle surface, an intraparticle diffusion coefficient D, characterizing diffusional transport along the interior carbon particle pore surfaces, and appropriate isotherm coefficients to characterize the functional dependence of the solid-phase concentration of adsorbed solute on the solution-phase concentration at equilibrium. In the modeling approach employed, these coefficients were evaluated for the target compounds directly in the presence of complex leachate material, which was considered as unspecified but system-specific background and quantified only in terms of total organic cqrbon (TOC). The rationale for this approach is that many field-scale applications of adsorption involve waters which are complex in composition and hssociated adsorbent-solute-solution interactions that are typically site specific, necessitating development of correspondingly site-specific design criteria. Further, the operation of fixed-bed adsorbers is often governed by the breakthrough pattern of one or two compounds, designated as “target” solutes either by virtue of their being the only identified hazardous substances in the treatment stream or because they are the first to exceed prescribed effluent limits in the adsorber system. Such an approach is easier to implement and affords reduced model input and computational requirements over methodologies which seek to identify and predict the adsorptive behavior of all components present. Data Collection and Analysis
A two-stage experimental program was implemented to obtain equilibrium and mass-transport coefficients for a matrix of two target organic compounds and background waters containing various amounts of DOM measured as TOC. In phase I, batch-mode isotherm and rate studies were conducted for one- and two-component solutions of the target compounds. Phase IT experiments utilized information from phase I to formulate fixed-bed adsorber experiments: short-bed adsorber (SBA) studies as an alternate method for kinetic parameter estimation and
0 1988 American Chemical Society
Environ. Sci. Technol., Vol. 22, No. 3, 1988 313
model calibration and deep-bed adsorber (DBA) experiments for parameter/model verification. Activated Carbon. The adsorbent used in all experiments was Filtrasorb 400 activated carbon (F-400, Calgon Corp., Pittsburgh, PA). The general physical properties of F-400 are documented elsewhere (13). For isotherm studies, the carbon was utilized in powdered form to facilitate rapid attainment of equilibrium and limit interferences due to biological activity. Powdered carbon was prepared by crushing random bag samples and saving the 200/325 sieve size fraction. For batch rate and fixed-bed adsorber experiments, the 30/40 sieve size fraction was used. Upon sieving, the carbon was washed with deionized-distilled water, oven-dried at 104 "C, and stored in air-tight glass containers. Carbon for immediate use was dried to a constant weight and then cooled at ambient temperature in a desiccator. Solutes. Two organic compounds, trichloroethylene (TCE) and p-dichlorobenzene (p-DCB), both designated as priority pollutants by the US.EPA, were chosen as target solutes. These compounds were selected because they exhibit different adsorption characteristics, have been identified in contaminated surface waters and groundwaters, are relatively straightforward to analyze, and embody a wide range of properties and characteristics. The majority of multicomponent adsorption studies to date have involved compounds from similar organic class groupings, whereas the intent here was to examine compounds having different structural and solution characteristics. TCE is a straight-chain, unsaturated aliphatic of relatively high volatility and solubility compared to the aromatic p-DCB. Background DOM consisted of a leachate collected from a hazardous landfill cell (HWL, Wayne Disposal, Rawsonville, MI). Raw leachate was prefiltered through a glass fiber filter prior to use to remove suspended particles that might adsorb pollutants in competition with activated carbon. The leachate was characterized as a high-strength waste (TOC = 10000 mg/L; total hardness = 2100 mg/L as CaC03), and its color and adsorptive characteristics suggest the presence of significant amounts of humic material. Appropriate amounts of HWL were diluted with buffered deionized-distilled water to achieve background DOM concentrations of 0,16,60, and 200 mg/L as TOC. Working solutions consisted of background water spiked with TCE and/or p-DCB, prepared as methanol-based stock solutions, to the desired concentration (a slight variation of this procedure employed for application of TCE to fixed-bed adsorbers is noted in the section on column methods). Experiments were conducted at room temperature (22 f 2 "C) and all solutions buffered at pH M phosphate. 6.5 f 0.2 with Mass concentration determinations for TCE and p-DCB were by gas chromatography using a liquid-liquid extraction procedure for sample preparation and an external standard calibration procedure. Background organic analysis was performed by direct TOC measurement or by ultraviolet spectroscopy correlated with TOC measurements. Isotherm Studies. Equilibrium data were collected according to the completely mixed batch reactor (CMBR) bottle-point technique. Varying amounts of powdered carbon were carefully measured and added to a series of 0.16-L glass vials, followed by contact with solutions containing the adsorbates. Pre1iminar.y investigations indicated that virtual equilibrium could be obtained for the solutes selected for study, including the leachate material, within a &day reaction period when powdered carbon was used. After the contact period, a sample was taken from 314
Environ. Sci. Technol., Vol. 22, No. 3, 1988
each reactor, filtered through a prewashed glass fiber filter to separate the carbon, and extracted or diluted according to the appropriate analytical technique. Upon measurement of the equilibrium solute concentration C,, the corresponding equilibrium solid-phase concentration q, was calculated from a mass balance. Control measures utilized in capacity experimentsincluded elimination of head space in reactors to limit volatilization losses and evaluation and accomodation of losses encountered in the filtration step. For single-solute systems of p-DCB and TCE, the Freundlich isotherm model was found to adequately describe the equilibrium liquid-phase/solid-phase relationship over the concentration ranges of interest. The Freundlich equation is a semiempirical, nonlinear expression of the form qe = K F C ~ ~
(1) Freundlich model coefficients were determined according to a nonlinear geometric mean functional regression algorithm which recognizes errors encountered in measurement/calculation of both liquid- and solid-phase equilibrium concentrations (14, 15). Multicomponent adsorption equilibria were described by Ideal Adsorbed Solution Theory (IAST), with an empirical modification similar to that employed by others (5, 16) to provide a more precise fit of the data. The generalized equations to be solved for the modified IAST are K , = ?r (2) N
czi = 1
i=l i
(3)
N
(4) 4, = ZiqT = (COJ- C,)/(M/V)
(5)
qi* = f ( C i * )
(6)
c, = P,Z,C,*
(7) where Piis incorporated as an empirical coefficient to account for competitive interactions between the target solutes in a specific system. When eq 1 is used to characterize f(Ci*) and to evaluate the spreading pressure ~ i eq 2-7 reduce to the expression
where g is the inverse of the Freundlich exponent n. The competition coefficient values, P,, are searched from bisolute equilibrium data with a minimization procedure. Equation 8 can be solved numerically by implementing a Newton-Raphson algorithm. A detailed development of the solution is given elsewhere (15). Kinetic Studies. Two techniques for mass transport parameter estimation were compared. Following a more traditional approach, CMBR rate experiments were used in conjunction with a batch-reactor MADAM algorithm to obtain D,from time/concentration data. The reactor had a volume of 2.3 L and was sealed and head-space free to prevent volatilization losses. Adequate mixing was provided to eliminate the effects of liquid-phase mass transfer, as established by exceeding minimum values of the Biot number calculated as described elsewhere (17). Initial target compound and background leachate concentrations approximated the influent concentrations used in subsequent column experiments.
,
Table I. Freundlich Isotherm Coefficients for Target Compounds in Single-Solute Systems" solute
background
p-DCB
HWL(0) HWL(16) HWL(6O) HWL(2OO) HWL(0) HWL(16) HWL(6O) HWL(2OO)
TCE
KF 31.55 27.07 14.43 8.15 1.61 1.37 0.79 0.43
95% confidence interval (KF)
n
95% confidence interval (n)
correlation coefficient
28.91-34.43 23.00-31.86 11.62-17.92 6.45-10.33 1.51-1.72 1.27-1.49 0.73-0.86 0.38-0.49
0.330 0.323 0.374 0.368 0.515 0.435 0.398 0.391
0.316-0.344 0.296-0.349 0.333-0.415 0.333-0.403 0.499-0.531 0.4 18-0.45 2 0.378-0.417 0.368-0.414
0.996 0.995 0.991 0.995 0.998 0.999 0.998 0.998
'
aBased on C. in x / L and 9, in a/m?z.
Film diffusion coefficients for corresponding fixed beds are typically estimated from literature correlations when the technique outlined above is used. In an alternative approach described here, a fixed-bed adsorber version of MADAM was calibrated with SBA data to simultaneously determine kfand D,.In the particular experimental design employed, the SBA, defined as a bed of sufficiently short length that immediate concentration breakthrough occurs (18), was placed in series with a second adsorber, the combined depths of the two columns comprising a DBA. The merit of this experimental design is that identical influent flow and concentration are assured for complementary short and long column runs for the purposes of model calibration and verification, respectively. The columns were constructed of borosilicate glass with an i.d. of 1.3 cm. For 30/40-mesh carbon, this gives'a columnto-particle diameter ratio of 25, essentially eliminating hydrodynamically related wall effects (19). Columns were packed with successive layers of micro glass beads (of the same size as the carbon), GAC, and more beads to establish a consistent flow pattern in and out of the bed. Stainless steel mesh screens were placed at the entrance and exit of the columns to retain the media. The beds were carefully packed in distilled water to eliminate air and then rinsed for several minutes at higher than the design flow rate to wash out remaining fines. Bulk influent solutions to the columns, which were operated in upflow mode, were prepared in and delivered from a 45-L glass container by a variable-speed peristaltic pump. To avoid volatilization, TCE was pumped to the system from a concentrated stock solution prepared in a well-mixed, head-space free cylinder. A stirred 1.2-L glass chamber with baffles was inserted upstream of the influent to the SBA to provide adequate mixing of incoming volatile compound stock with the bulk solution. An air trap was employed between the bulk feed pump and mixing chamber to prevent air from passing to the carbon beds. All tubing and valves were made of glass to minimize adsorption of organics onto reactor surfaces. The system was fitted with sampling ports before and after column 1and after column 2 for influent and SBA and DBA effluent sampling, respectively. Samples were collected at discrete time intervals and extracted immediately for analysis. Results and Discussion
Equilibrium Studies. A tabulation of Freundlich isotherm coefficients for single-solute solutions of p-DCB and TCE in the presence of various background concentrations of leachate is given in Table I. Experimental data and Freundlich model plots for TCE and p-DCB are presented in log-log format in Figures 1and 2, respectively. The results indicate that the capacity of F-400 for p-DCB is much greater than for TCE and that the presence of background DOM reduces the adsorption capacity for both target compounds. Moreover, the decrease in adsorption
TOC
TOC TOC tOC I
h d
h&
hX)'
h10'
EQUlLlBRlUM CONCENTRATION (uG/L)
Flgure 1. Single-solute isotherms for TCE in the presence of varying concentrations of leachate DOM.
P I hO1
I
I
b&
,
,
,
. , . . .I I
.. . '. .. I
bld
EQUILIBRIUM CONCENTRATION (UG/L)
Figure 2. Singlasolute isotherms for p-DCB in the presence of varying concentrations of leachate DOM.
capacity is proportional to the background DOM concentration expressed in terms of TOC. These capacity effects are reflected in the magnitude of the Freundlich capacity factor KF. The capacities for both target compounds are affected to a similar degree in the leachate background; e.g., for HWL(GO), adsorption capacities are reduced to 45-50% of the values for the case of no background DOM. [Note: HWL(X) denotes hazardous waste leachate background with an initial concentration of X mg/L as TOC.] The slope terms, n, for the p-DCB isotherms exhibit little variation, a result consistent with other studies conducted for target compounds in background DOM (1,6). In the case of TCE, however, n values in leachate background are approximately 20% lower than for the case of no background DOM. Due to the semiempirical nature of the Freundlich model, it is difficult to attach absolute physical significance to the observed behavior, although the slope term is roughly an indicator of adsorption energy or intensity (20). As illustrated in Figure 1,this translates for this system to TCE adsorbing relatively more favorably Environ. Scl. Technol., Vol. 22, No. 3, 1988
315
Table 11. Competition Factors, Pi,for Bisolute Isotherm Model for Various Background Leachate Concentrations
c,, UBIL
PL
background
p-DCB
TCE
p-DCB
HWL(0) HWL(0)
4127.0 2868.0 5857.0 5774.0 4813.0 1258.0 5776.0 5899.0
843.0 1115.0 1130.0 1897.0 969.0 966.0 972.0 1860.0
1.23 1.21 1.24 1.27 1.15 1.14 1.09 1.06
HWL(0) HWL(0) HWL(16) HWL(6O) HWL(6O) HWL(6O)
av P,
TCE 0.73 0.71 0.73 0.74 0.22 0.16 0.16 0.13
pDCB
U""
TCE
una
1.24
0.02
0.73
0.01
1.15 1.10
0.22 0.03
0.15
0.01
"Standard deviation based on N events.
a t the very low end of the concentration range than a t higher concentrations when HWL is present. The precise mechanisms and solid/solution characteristics responsible for reduction by DOM of target compound adsorption are not readily apparent. Aquatic humics have demonstrated a tendency to form complexes with certain organic micropollutants, altering such solution properties as solubility and hydrophobicity (21-23) and, therefore, adsorption characteristics ( 2 , 6 ) . In this context it is relevant to note that solution preparation procedures for the isotherm experiments allowed only a few minutes between injection of concentrated target solute into background DOM and contact with activated carbon. Thus, if the association of a compound with humic material is rate-limited, the results presented may not reflect the total magnitude of potential complexation effects. Direct competition for adsorption sites between target compounds and the strongly adsorbing components of the background DOM is certainly another potential mechanism for reduction of the apparent adsorption capacity of the former. A related factor is the presence of detectable levels of Ca2+ and other inorganic salts in the leachate, substances known to enhance the adsorption of humics on activated carbon (24). Physical and chemical properties specific to a particular activated carbon can also play determinative roles in the nature of adsorption of various types of DOM. It is evident then that more precise characterization of the background DOM and investigation of isolated solute/ DOM/adsorbent interactions and their relative contributions to observed adsorption phenomena are required to identify operative mechanisms. Data for TCE for two-component isotherms in various background leachate concentrations are shown in Figure 3 along with modified IAS model calibrations (solid lines). Comparison of these data and model traces with the dashed lines, which represent single-solute isotherms for TCE in the presence of HWL(0) and HWL(16), illustrates the significant reduction in TCE adsorption capacity effected by the presence of p-DCB. The presence of leachate in solution further suppresses TCE adsorption. The displacement of the bisolute isotherms from their corresponding single-solute curves is greatest in the upper concentration range. In fact, the tail of the isotherm bends downward, a common feature of isotherms for competitive adsorptions between compounds which sorb strongly and weakly relative to each other. This region corresponds to very small carbon dosages and involves high surface coverage and more intense competitionfor available sites. The more strongly adsorbing solute, p-DCB, has a greater energy for adsorption than does TCE and consequently dominates surface coverage in this region. Computed competition coefficients, Pi,for each multisolute isotherm are presented in Table 11. Values of Pi varied significantly from 1.0 for both p-DCB and TCE for every background concentration of leachate studied. The 316
Envlron. Sci. Technol., Vol. 22,
No. 3, 1988
5
0
t
at
0
200
400
600
800
1000
EQUILIBRIUM CONCENTRATION (UG/L)
-
-
Figure 3. Isotherms for TCE in the presence of p-DCB for various concentrations of leachate DOM: Co,TcE 1000 pglL; Co,p,, 5000 pg/L. Dashed lines are corresponding TCE single-solute isotherms.
term Piis essentially a measure of unequal or so-called nonideal competition between adsorbates. As a phenomenological coefficient, it also accounts for any hysteresis = effects on adsorption equilibria. Ideal competition (Pi 1.0) implies that the solutes adsorb "independently" in the sense that there are no solute-solute, solute-DOM, or solute-sorbent interactions or energetics operative other than those manifest in their respective single-solute isotherm patterns and the stoichiometry of the bisolute system. Inherent also is the assumption of a homogeneous adsorbent with respect to active sites. This assumption is not met by most activated carbons, the majority of which possess numerous functional groups as well as variable pore size distribution, the nature of which can promote partial or total steric exclusions of certain solutes (5). It is difficult to correlate the nonidealities reflected by the competition coefficient values with specific system parameters to make a priori estimates of adsorption equilibria and thereby circumvent the need for multicomponent isotherm data. That PTCE values are less than 1.0 and that P .DCB values are greater than 1.0 indicate that, for the stateiconditions, TCE competes more favorably with p-DCB for adsorption on F-400 than predicted by ideal competition. It is important to recognize several limitations to the modeling approach applied. First, the competition coefficients are valid over only a limited initial concentration range. Moreover, Pimust approach 1.0 as the concentration of one (or both) of the sorbates becomes very small (e.g., the model must predict single-solute equilibria when only one solute is present). In addition, the IAST formulation used in this study is valid because the singlesolute isotherm data were well fitted by the Freundlich model. If significant curvature exists in the log-log trace of single-solute equilibrium data, use of the Freundlich equation in the IAST solution will result in substantial errors in calculation of the spreading pressure. In such
-8
t
0 0
0
;ss
NO BACKGROUND HWL HWL AS TOC
o 60 MG/L
-0
Table 111. Comparison of Film Mass Transport Coefficients
.
p-DCB in HWL(60)'
4,
method of determination
i "000 000
8 000 0
1600
T
&'EO"( HO u3Rzsop
4000
4800
I
5600
Flgure 4. CMBR rate data and model calibration profiles for p-DCB in HWL(0) and HWL(6O) backgrounds.
Sh cm/s
experimental (SBA) Williamson et al. (25) Wilson and Geankoplis (26) Gnielinski (27) Ohashi et al. (28) Kataoka et al. (19) Dwivedi-Upadhyay (29)
X
TCE in HWL(60)b lo3 Sh
5.8 6.6 7.7 7.5 5.6 8.1 8.1
39.6 46.1 45.1 34.0 48.5 49.0
36.8 43.2 42.5 32.1 45.5 45.9
ki,
cm/s
X
IO3
5.0 7.3 8.6 8.5 6.4 9.1 9.1
"Re = 10.4; Sc = 1116.3. bRe = 10.3; Sc = 931.4.
Table IV. Rate Coefficients for Single-Solute Systems
kf,cm/s x solute p-DCB
TCE
back: ground none HWL(6O) none
HWL(6O)
io3
SBA
cor (best/worst)'
7.9 5.8 6.6 5.0
8.0/5.9 5.6/8.1 6.3/9.0 6.4/9.1
D,, cm2/s X 1010 SBA CMBR 1.60 0.57 5.60 4.20
2.90 2.60 10.20 5.90
Best and worst based on deviation from kt (SBA).
'bo0
2000
4000
6000
8000
10000
12000
14000
BED VOLUflES ( I N THOUSANDS)
Flgure 5. SBA data and model calibration for p-DCB In HWL(6O) background.
situations, a multiparameter isotherm model which better describes (log-log) nonlinearity must be applied. Kinetic Studies. The questionable accuracy of model inputs which must be obtained experimentally or from empirical correlations is always a concern for modeling efforts employing dynamic models of relatively high levels of mathematical sophistication. This is particularly true relative to mass-transport parameter evaluation for multiple-resistance adsorption models. Such evaluations are commonly done by subjecting CMBR rate data to a statistical parameter search for estimation of the k,, (Le., kf for CMBR conditions) and D, values. CMBR rate data and MADAM batch model calibrations for p-DCB in HWL(0) and HWL(6O) are illustrated in Figure 4. Because the hydrodynamic characteristics of a fixed-bed reactor are significantly different from those of a CMBR, correlation techniques are used to estimate kf values for column modeling. Associated errors in both search and correlation techniques can result in the compounding of common-direction errors. The SBA technique used here more closely approximates the hydrodynamics of full-scale columns than does a CMBR, allows simultaneous determination of kf and D,, and minimizes error compounding by mutual compensation of errors in kf and D, during parameter searchlregression analysis. Figure 5 shows typical SBA data and MADAM calibration results for p-DCB as a single target solute in HWL(60) background. Film transport is assumed to control at the initial breakthrough stage. The value of kf is searched with a single-compound column version of MADAM to fit the first 20-30 min of data, followed by calibration of D, over the entire profile. Both -search routines utilize a minimization function based on the sum of the squares of the errors between data and model calculations. Desorption data collected after reduction of the influent concentration of p-DCB are included in Figure
5. The model calculations for this portion of the profile are based on an assumption of complete reversibility; Le., equilibrium and rate parameters are those obtained from the adsorption region. Numerous film mass transport correlations are reported in the literature, each distinguished by a particular functional relationship between the dimensionless Reynolds (Re) and Schmidt (Sc) numbers in the expression for the Sherwood number (Sh), which is related to kf by Sh = k&/D, (9) Re and Sc are defined as Re = v , d / v (10) SC= v/DL (11) Sample kf values for single-solute cases of p-DCB and TCE in HWL(6O) computed from six different correlations are presented in Table 111. Separate calculation of kf is required for each experiment because film diffusion is a function of flow conditions and bed void fraction. Bulk liquid diffusivities, DL,for p-DCB and TCE were determined with the Wilke-Chang equation (30). Table I11 also includes kf values estimated by the SBAIMADAM calibration technique. The different correlation values for kf may be observed to vary over a significant range, with most differing substantially from those obtained with the SBA experimental technique. Mass transport coefficients for single-solute solutions of p-DCB and TCE in leachate background from both parameter determination techniques are given in Table IV. The SBA calibrations of kf and D,and CMBR determinations of D,reflect the decreased rates of film and intraparticle diffusion in the presence of background DOM observed in the data. Literature correlation calculations of kfdo not account for interactions between target compounds and therefore cannot reflect the rate decreases. In addition, the correlation equations presented in Table I11 have been developed with materials that are substantially different in chemical and physical character than microporous adsorbents such as activated carbon. A number of studies have demonstrated that the surface topography and roughness of an adsorbent can impact film-controlled Environ. Sci. Technol., Vol. 22, No. 3, 1988
317
1 -SEA 2 - Corr(E)-CMER
a 2-1
1-SEA 2 - Corr(E)-CMER 2-Corr(E)-CMER 3-Corr(W)-CMBR
a/
m
00 l 0 ”
tom
15.m
x.m
M.m
a.m
BED VOLUMES ( I N THOUSANDS)
I
35.m
IL
rim
6:m
i m
2i.m
4.m
BED VOLUflES I I N THOUSANDS)
4.m
Figure 6. DBA data and model predictions for p-DCB in HWL(6O)
Flgure 7.
background.
background.
Table V. Physical Parameters Used for Single-Solute DBA/MADAM Model Verification Studies
Table VI. Comparison of Breakthrough Profiles for Single-Solute Systems Generated by Different Rate Parameter Determination Techniques [Background HWL(6O)I
solute
background
C,, wg/L
e
u,, cm/s
L,cm
p-DCB
HWL(0) HWL(6O) HWL(0) HWL(6O)
3484 3352 641 660
0.356 0.391 0.399 0.395
2.11 1.93 1.89 1.92
4.5 6.1 5.6 6.0
TCE
mass-transfer rates (27,31). Such impacts may vary according tQ the nature of the solute and be augmented when competitive species are present. Moreover, no standard criteria exist for a priori selection of a particular correlation for a specific system. Table IV lists “best” and “worst” correlation values, where best and worst are based upon deviations from kfdetermined from the SBA approach. In the case of p-DCB, correlation kf values bracketed the experimentally determined coefficients. For TCE, however, correlation values were usually higher than SBA values, especially when leachate was present in the background. Intraparticle diffusion coefficients estimated with the CMBR method are higher than those determined from the SBA for both compounds, the largest percentage deviation existing in the case of p-DCB in leachate background and the smallest for TCE in leachate background. The discrepancy in parameter values between the two techniques is assumed to be due to differences in the hydrodynamic conditions of the experimental systems, as well as to differences in the extent to which variations in these conditions are reflected in the corresponding kfvalues. For instance, the sequential loading of multiple adsorbing species is likely different in the two reactor configurations. It is anticipated that solid and/or solute interactions associated with the presence of additional sorbates, especially humic-like substances, may accentuate such differences, particularly in the case of intraparticle mass transport within microporous adsorbents. Both techniques predict that TCE diffuses more rapidly than p-DCB along the internal surfaces of F-400 whether background DOM is present or not. Figures 6 and 7 illustrate model verification studies for p-DCB and TCE, respectively, in HWL(60). Predicted MADAM profiles were generated with the physical parameters listed in Table V and (1) SBA rate parameters, (2) best correlation kfwith CMBR D,, and (3) worst correlation kf with CMBR D,. Lines 2 and 3 bound the profiles generated with kfvalues estimated from the six literature correlation procedures. Clearly, the SBA approach provides a better prediction of DBA breakthrough data than the correlation-CMBR profiles for the case of p-DCB. The SBA-based prediction is adequate over the entire profile, while the correlation-CMBR method sig318
Environ. Sci. Technol., Vol. 22, No. 3, 1988
DBA data and model predictions for TCE in HWL(6O)
%
solute p-DCB TCE
4i.m
breakthrough 10 25 50 10 25 50 75
bed volumes treated (in thousands) predicted cor(B)cor(W)measured SBA CMBR” CMBRb 1.8 7.1 12.2 0.2 1.2 2.5 6.9
3.7 8.5 12.8 0.5 1.4 2.6 7.0
4.6 16.6 32.0 0.9 1.7 3.1 8.4
9.1 18.7 32.0 1.2 1.9 3.1 8.3
Cor(B)-CMBR kf from best correlation; D from CMBR rate data. Cor(W)-CMBR: k f from worst correlation; D from CMBR rate data.
nificantly overestimates the performance of the column. This is depicted numerically in Table VI, which compares experimentally observed numbers of bed volumes treated to prescribed effluent levels to values predicted with models calibrated by the various parameter estimation techniques. All of the modeling approaches provide reasonably good prediction of the data in the case of TCE in leachate background, although the SBA technique gives a better fit to the initial portion of the breakthrough profile (see Figure 7 and Table VI). In this region, discrepancies between the data and model predictions are attributable to errors associated with the estimated value of kf. The differences in D, values obtained by the two calibration techniques are much greater for p-DCB than for TCE, resulting in the correspondingly greater discrepancies between MADAM profiles generated for p-DCB. Model predictions in the desorption are also more accurate for TCE than for p-DCB, suggesting that a significant portion of p-DCB adsorption may be irreversible. Another pattern observed throughout is that the deviation between actual and model desorption values is greater in the DBA (Figure 6) than in the SBA (Figure 5). Typical SBA data and MADAM calibrations for a bisolute mixture in leachate background are presented in Figure 8. The small overshoot of the TCE profile is evidence of a competitive effect exerted by the more strongly adsorbed compound, p-DCB, on the more weakly but more rapidly adsorbed TCE. This type of displacement phenomenon is a common characteristic of multicomponent systems in which the adsorbing species have substantially different energies and rates of adsorption. Table VI1 contains rate coefficients for the bisolute studies. Film mass transport values computed from lit-
Table VIII. Physical Parameters Used in DBA-Mode Verification Studies for Bisolute Systems
am
6.m am x.m BE0 VOLUtlES (IN THOUSRNOS)
am
am
am
a m
Figure 8. SBA data and model callbrations for p-DCB and TCE in leachate background.
no.
background
B1 B2 B3 B4 B5 B6
HWL(0) HWL(16) HWL(6O) HWL(6O) HWL(6O) HWL(6O)
background
no.
solute
B1
p-DCB TCE p-DCB TCE p-DCB TCE p-DCB TCE p-DCB TCE p-DCB TCE
B2 B3 B4 B5 B6
HWL(0)
7.7 6.7 7.5 5.3 5.3 3.3 5.4 3.3 5.7 3.7 3.9 2.4
HWL(l6) HWL(6O) HWL(6O) HWL(6O) HWL(6O)
%
D,, cm2/s X 10'0 SBA CMBR 1.8 20.0 0.62 16.0 0.37 9.2 0.36 11.0 0.30 9.8 0.36 9.8
3.0 57.0 2.3 52.0 2.1 32.0 2.1 32.0
3438 3537 3285 3425 751 3398
637 676 667 684 645 657
t
uz, cmjs
L , cm
0.388 0.370 0.399 0.386 0.371 0.396
1.94 2.03 1.89 1.94 2.02 0.94
7.0 7.1 7.0 11.0 6.0 8.0
Table IX. Comparison of Breakthrough Profiles for Bisolute Systems Generated by Different Rate Parameter Determination Techniques (Run B6) [Background HWL(16)I
Table VII. Rate Coefficients for Bisolute Systemsn kn cm/s X lo3, SBA
c,, U d L p-DCB TCE
solute p-DCB TCE
breakthrough 10 25 50 10 25 50 75
bed volumes treated (in thousands) predicted cor(B)cor(W)measured SBA CMBR" CMBR" 9.8 13.7 22.8 1.5 2.7 4.6 8.0
10.2 14.9 21.0 1.5 3.4 5.4 7.4
17.4 27.6 >32.0 2.6 5.0 7.5 9.6
14.1 26.5 >32.0 3.5 5.5 7.5 9.2
Cor(B)-CMBR: k f from best correlation; D from CMBR rate data. bCor(W)-CMBR kf from worst correlation; D from CMBR rate data. (I
2.1 32.0
See Table VI11 for corresponding bed porosity and velocity conditions.
erature correlations are essentially the same as those in single-solute systems as this approach assumes that the sorbates diffuse independently of one another. As noted earlier, however, competitive adsorption equilibria in the systems studied were observed to be of nonideal character, and it is reasonable to assume that adsorption kinetics are similarly affected. Observations regarding the rate parameters for the bisolute analysis are consistent with those for single-solute modeling; namely, (1)both kfand D, for both target compounds decrease with increasing background DOM concentrations, (2) values of kfobtained for TCE from literature correlations are significantly higher than those determined by the SBA in leachate backgrounds, (3) D, values obtained from CMBR data are higher than corresponding values obtained from SBA data, and (4)D, values for TCE are higher than D, values for p-DCB in each case. Comparison of the single-solute coefficients presented in Table IV to the bisolute coefficientsin Table VI1 reveals that D,values for p-DCB are similar, but the values determined for TCE in the bisolute systems are much higher than the corresponding single-solute value. This peculiarity is due, in part, to the nonideal equilibrium behavior of p-DCB and TCE in bisolute systems represented by the values of P T Cbeing ~ appreciably less than 1.0. This nonideality is reflected in the intraparticle rate coefficients in response to a numerical feature of MADAM relating to the value of the solute distribution parameter D . The solute distribution parameter, defined in the MADAM algorithm as N
N
D, = P ( 1 - 4C40,L/(ECCo,i) i=l
1=1
(12)
Flgure 9. DBA data and model predictions for p-DCB and TCE in HWL( 16) background.
is utilized in the expression that dedimensionalizesa term which includes D, for solution of the solid-phase material balance equation. Thus abrupt changes in D,may result in a similar discontinuity in the same direction for D, for variable calibration runs. Initial experimental conditions and relative equilibrium partitioning of p-DCB and TCE are such that D, values in the bisolute system are approximately 5 times higher than the D,values in singlesolute calibrations for TCE, resulting in the corresponding shift in D, values determined by MADAM (6). Physical parameters for the various bisolute DBA runs are listed in Table VIII. Comparisons of model predictions to DBA data for the sample bisolute run described by the information given in Figure 9 and Table IX yield conclusions which closely parallel those obtained from comparisons of single-solute data and model predictions. Breakthrough curves determined with the SBA technique predict DBA data much more closely than correlationCMBR procedures for p-DCB, with model deviations most attributable to variations in D,. Prediction of TCE breakthrough is not as sensitive to the parameter estiEnviron. Sci. Technol., Vol. 22, No. 3, 1988
319
--
LINE
WL C O N (AS TOC)
1
0 mq/l 16 mg/l 60 mg/l
2
3
Vm
s'm
dm
BED ?O?UMES TN~THo&SA~NDS)
som
am
com
Flgure 10. Model DBA breakthrough profiles for TCE in the presence of p-DCB for varying concentrations of background HWL.
Flgure 11. Model DBA breakthrough profiles for p-DCB in the presence of TCE for varying concentrations of background HWL.
mation methodology used, although the SBA technique is more accurate in the earlier portion of the profile and provides a better overall prediction of DBA behavior. The competitive displacement of TCE is more evident in the DBA than in the SBA profiles. Although the extent of TCE desorption is not predicted precisely by MADAM, the general pattern is obtained, particularly when SBAestimated coefficients are used. Examining the regions bounded below and above the C/C, = 1.0 line, it is apparent that a large fraction of the TCE adsorbed initially is displaced into the effluent by the more strongly adsorbing p-DCB. This phenomenon is further evident in Figure 10, which shows MADAM profiles for TCE in bisolute systems for varying concentrations of leachate background (based on SBA rate coefficients). Figure 11 presents corresponding profiles for p-DCB. These several examples illustrate the decreases in adsorber performance observed with increasing background DOM concentrations. The combined impacts of reduced equilibrium capacities and rates on adsorber performance is significant at leachate concentrations as low as 16 mg/L as TOC.
Summary GAC adsorption of two target organic compounds in the presence af background DOM comprised of a hazardous waste landfill leachate was analyzed by an existing mathematical adsorption model, MADAM. The modeling approach applied was one in which model parameters were experimentally determined and calibrated for the targt compounds only, with the leachate considered as an unspecified background. Findings and conclusions deriving from the study are summarized below. (1)The presence of leachate DOM reduced adsorption capacities and rates for both target compounds (p-DCB and TCE) compared to their respective values in water having no background DOM. These impacts are reflected 320
Environ. Sci. Technol., Voi. 22, No. 3, 1988
quantitatively in equilibrium and kinetic model coefficients determined by calibration of bench-scale experimentaldata and are proportional to the strength of the background water as quantified by its TOC content. (2) The SBA technique was verified as a useful method for determination of system-specific kinetic parameters for complex mixtures of organics. Film and surface diffusion coefficients estimated by this method resulted in adequate predictions of the performance of fixed beds for single and multicomponent cases for the range of system conditions studied. (3) A more traditional methodology for estimating rate parameters utilizing literature correlations and CMBR rate data did not predict fixed-bed adsorber breakthrough profiles as accurately as the SBA technique. Discrepancies resulting from application of the correlation-CMBR technique are more pronounced for p-DCB than for TCE and most attributable to differences in values for the internal diffusion coefficient. Values of D,determined from CMBR rate data were significantly higher than those estimated from SBA analysis, presumably due to the different hydrodynamic conditions prevailing in the two reactor systems and to differences in the extent to which differences in these conditions were reflected in the corresponding kf values. Overprediction of fixed-bed adsorber performances resulted when the correlation-CMBR parameters were used in the MADAM model projections. Differences in film mass transport coefficient values for fixed-bed adsorbers result from the fact that literature correlations do not account for interactions between target compounds and background DOM, while these interactions appear to be reflected in kf values derived by the SBA technique. While the proposed modeling approach has been shown to apply reasonably well for the cases investigated, a deeper and more fundamental understanding of the impacts of DOM on mass transfer in fixed-bed adsorbers is required to evaluate the merits of extrapolating the methodology to other systems. From a process viewpoint, the interpretation and value of the results could potentially be enhanced by further research using carbons of variable particle size and pore size distribution. Consideration should also be given to refining the MADAM-HSD algorithm to eliminate possible bias in the generation of mass-transfer parameters such as those apparently effected by the dimensionless solute distribution parameter. Despite these potential limitations, the reduced input and computational requirements of the modeling methodology described herein make it an attractive prospect as a practical design tool for applications in which primary concern relates to the removal of only one or a few known target organic compounds from complex mixtures. Acknowledgments We express our appreciation to Brett Farver, Barbara Jacobs, and Daniel Peters for their contributions to the experimental aspects of this project. Glossary
ce Ci
c,*
equilibrium liquid-phase solute concentration Gg/L) liquid-phase concentration of species i, IAST (pM) single-solute concentration of species i evaluated at the bisolute mixture spreading pressure, IAS (FM)
dC"4
initial concentration of species i(pM or pg/L) carbonsolute particle diameter (cm)
D,
solute distribution parameter (dimensionless)
bulk liquid diffusivity (cm2/s) surface diffusion coefficient (cm2/s) D, internal diameter (cm) i.d. Freundlich isotherm constant KF film diffusion coefficient (cm/s) kf carbon bed depth (cm) L carbon dose (mg of adsorbent) M number of target species in solution N n Freundlich isotherm exponent (dimensionless) empirical competition coefficient for species i, IAST pi (dimensionless) equilibrium soolid-phase solute concentration 4e (fig/mg) solid-Dhase concentration of species i, IAST 4i (fimol/mg) solid-phase concentration of species i in single-so4i* lute system evaluated at the bisolute mixture spreading pressure, IAST (pmol/mg) initial solid-phase concentration of species i40,i (mol/mg or fig/mg) total solid-phase concentration of solutes, IAST 4T (fimol/mg) liquid volume in reactor (L) V interstitial flux velocity in carbon bed (cm/s) vz solid-phase mole fraction of species i, IAST (dizi mensionless) bed void fraction (dimensionless) 6 inverse of Freundlich exponent n (dimensionless) II kinematic viscosity (cm2/s) V spreading pressure, IAST (kcal/cm2) a carbon particle density (g/cm3) P Registry No. Cl,C=CHCl, 79-01-6; 4-C1C6H4Cl,106-46-7. DL
Literature Cited Weber, W. J., Jr.; Pirbazari, M. J.-Am. Water Works Assoc. 1982, 74, 203. Frick, B. R.; Sontheimer, H. In Treatment of Water by Granular Activated Carbon;Suffet, I. H., McGuire, M. J.,
Eds.; American Chemical Society: Washington,DC, 1983;
247 p.
Weber, W. J., Jr.; Liang, S. Environ. Prog. 1983, 2, 167. Callaway,J. Y.; Gabbita, K. V.; Vilker, V. L. Environ. Sci. Technol. 1984, 18, 890.
Yonge, D. R.; Keinath, T. M. J.-Water Pollut. Control Fed. 1986, 58, 77.
Smith, E. H.; Tseng, S.; Weber, W. J., Jr, Environ. Prog.
(9) Endicott, D. D.; Weber, W. J., Jr. Environ. Prog. 1985,4, 105. (10) Crittenden, J. C.; Luft, P.; Hand, D. W. Water Res. 1985, 19, 1537. (11) Crittenden, J. C.; Weber, W. J., Jr. J. Environ. Eng. Diu. (Am. SOC.Ciu. Eng.) 1978, 104, 1175. (12) Weber, W. J., Jr. In Proceedings, First International
(13) (14) (15) (16)
Conferenceon the Fundamentals of Adsorption; Myers, A. L., Belfort, G., Eds.; Engineering Foundation and AIChE: New York, 1984; 647 p. Wang, C. Ph.D. Dissertation, University of Michigan, Ann Arbor, MI, 1986. Halfon, E. Environ. Sci. Technol. 1985, 19, 747. Smith, E. H. Ph.D. Dissertation, University of Michigan, Ann Arbor, MI, 1987. Thacker,W. E.; Crittenden,J. C.; Snoeyink, V. L. J.-Water
Pollut. Control Fed. 1984,56, 243. (17) Hand, D. W.; Crittenden, J. C.; Thacker, W. E. J. Environ. Eng. (N.Y.) 1983, 109, 82. (18) Weber, W. J., Jr.; Liu, K. T. Chem. Eng. Commun. 1980, 6, 49. (19) Kataoka, T.; Yoshida, H.; Ueyama, K. J. Chem. Eng. Jpn. 1972, 5, 132. (20) Weber, W. J., Jr. Physicochemical Processes for Water Quality Control; Wiley: New York, 1972. (21) Wershaw, R. L.; Burcar, P. J.; Goldberg, M. C. Environ. Sci. Technol. 1969, 3, 271. (22) Carter, C. W.; Suffet, I. H. Environ. Sci. Technol. 1982,16, 735. (23) Cdloway, J. Y.; Gabbita, K. V.; Vilker, V. L. Environ. Sci. Technol. 1982, 18, 890. (24) Weber, W. J., Jr.; Voice, T. C.; Jodellah, A. J. J.-Am. Water Works Assoc. 1983, 75, 612. (25) Williamson,J. E.; Bazaire, K. E.; Geankoplis, C. J. Ind. Eng. Chem. Fundam. 1963,2,126. (26) Wilson, E. J.; Geankoplis,C. J. Ind. Eng. Chem. Fundam. 1966, 5, 9. (27) Roberts, P. V.; Cornel, P.; Summers,R. S. J . Environ. Eng. (N.Y.) 1985, 111, 891. (28) Ohashi, H.; Sugawara, T.; Kikuchi, K. I.; Konno, H. J. Chem. Eng. Jpn. 1981, 14,433. (29) Dwivedi, P. N.; Upadhyay, S. N. Ind. Eng. Chem. Process Des. Dev. 1977, 16, 157. (30) Wilke, C. R.; Chang, P. J. AIChE J. 1955, I, 264. (31) Van Vliet, B. M.; Weber, W. J., Jr. J.-Water Pollut. Control Fed. 1981, 53, 1585.
1987, 6, 18.
Weber, W. J., Jr.; Smith, E. H. Environ. Sci. Technol. 1987, 21, 1040. Tien, C. In Proceedings, First International Conference on the Fundamentals of Adsorption;Myers, A. L., Belfort, G., Eds.; EngineeringFoundation and AIChE: New York, 1984; 647 p.
Received for review February 13,1987. Accepted September 23, 1987. The work was supported in part by Award CEE-8112945 from the National Science Foundation. The contents do not necessarily reflect the views and policies o f the NSF, and the mention o f trade names or commercial products does not constitute endorsement.
Environ. Sci. Technol., Vol. 22,
No. 3, 1988
321
