PREDICTION AND VERIFICATION OF ATRAZINE
A D S O R P T I O N BY P A C

Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 10/23/15. Copyright ASCE. For personal use only; all rights reserved.

By Shaoying Qi, 1 Samer S. Adham, 2 Vernon L. Snoeyink, 3 and Ben W. Lykins Jr. 4

ABSTRACT: A procedure was developed to predict the removal of trace organic compounds from natural water by powdered activated carbon (PAC) adsorption systems, which function as a batch reactor or a continuous stirred tank reactor
(CSTR). The procedure uses the equivalent background compound method coupled with the ideal adsorbed solution theory to quantify the competition between trace organicsand background organic matter in water, and uses the pseudo single-solute homogeneous surface diffusion model to describe the adsorption kinetics of the target compound under the influence of the background organic matter. The parameters required by the model as input data can be independently determined from adsorption isotherms and a set of batch kinetic test data. Good agreement between predicted and actual performance was found for adsorption of atrazine from Central Illinois ground water at different initial concentrations and different carbon doses using a batch reactor and two CSTRs, one of which was a PAC/ ultrafiltration system.
INTRODUCTION

Strict drinking-water standards have been p r o p o s e d for many organic compounds [Federal Register 1987, 52(No. 130; July 8), 1989, 54(No. 97;
May 22); Clark 1989], and the water industry needs to have an effective, economical way to meet the standards. P o w d e r e d activated carbon ( P A C ) adsorption is a promising means of controlling organic contamination in drinking water. Previous studies showed that activated carbon had high adsorption capacities for m a n y compounds (Miltner et al. 1989; N a j m et al.
1991a, 1990b; Speth and Miltner 1980). H o w e v e r , if the adsorption system is not properly designed and o p e r a t e d , the full adsorption capacity of the
P A C may not be achieved and the system may not be cost-effective. A n effective procedure is n e e d e d to evaluate system p e r f o r m a n c e in o r d e r to determine where changes in the process should be m a d e to improve adsorption efficiency. T h e purpose of this p a p e r is to present such a p r o c e d u r e and to apply it to the removal o f atrazine, a pesticide c o m m o n l y found in surface water and ground water.
BACKGROUND

Predictive methods for evaluating the performance of a P A C adsorption system involve the use of either equilibrium models or mass transfer models.
Both types of models in literature are usually based on the assumption that
P A C adsorption systems can be treated as a continuous stirred tank reactor
~Res. Asst., Dept. of Civ. Engrg., Univ. of Illinois, Urbana, IL 61801.
2Res. Asst., Dept. of Civ. Engrg., Univ. of Illinois, Urbana, IL.
3Prof., Dept. of Civ. Engrg., Univ. of Illinois, Urbana, IL.
4Chf., System and Field Evaluation Branch, U.S. EPA, Cincinnati, OH 45286.
Note. Discussion open until July 1, 1994. To extend the closing date one month, a written request must be filed with the ASCE Manager of Journals. The manuscript for this paper was submitted for review and possible publication on September 23,
1991. This paper is part of the Journal of Environmental Engineering, Vol. 120, No.
1, January/February, 1994. 9
ISSN 0733-9372/94/0001-0202/$1.00 + $.15 per page. Paper No. 2739.
202

J. Environ. Eng., 1994, 120(1): 202-218

(CSTR) or a plug flow reactor (Sontheimer et al. 1989; Najm et al. 1990a,
1990b).

Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 10/23/15. Copyright ASCE. For personal use only; all rights reserved.

Equilibrium Models
Sontheimer et al. (1989) derived an equilibrium model by incorporating only the adsorption isotherm into the mass balance equation for a CSTR.
The Freundlich isotherm was used for single-solute adsorption and the ideal adsorbed solution theory (IAST) of Radke and Prausnitz (1972) was used for multisolute adsorption. The authors indicated that the model could be used to determine the minimum carbon dose needed to reduce the pollutant concentration to the required effluent concentration.
Background organic matter in natural water strongly competes with target organic compounds for adsorption sites on activated carbon. As a result, the adsorption capacity for a target compound is often far below that indicated by the single-solute isotherm in distilled water. Good estimates of the amount of competition are needed in order to obtain reasonable estimates of PAC performance. Unfortunately, competitive adsorption models such as the IAST are not directly applicable to the problem because background organic matter is a complex mixture of many compounds with different adsorption characteristics. Najm et al. (1990b, 1991b) developed the equivalent background compound (EBC) method to calculate the capacity of PAC for a trace organic compound in the presence of background organic matter. The background organic matter was represented by a single compound, termed the EBC, and the competition between the EBC and the target compound was modeled by the IAST. [The approach is similar to that of Sontheimer et al. (1989) who assumed that the background organic matter could be represented by a number of fictive compounds.] The adsorption parameters (the initial concentration, C2.0, and the Freundlich isotherm parameters, K2 and l/n2) of the EBC were then determined by solving the two-componen} IAST equations using the isotherms for the target compound in distilled water and in the natural water as input data.
Thus, the EBC parameters are for a single compound that will produce the amount of competition observed by comparing the distilled-water and natural-water isotherms of the target compound. The EBC parameters and the distilled-water isotherm parameters can then be used in the IAST to calculate the isotherms for different initial concentrations of the target compound in the water with the same background organic matter. Najm et al. (1990b,
1991b) showed that the EBC method could be used to predict the decreasing capacity of activated carbon for trichlorophenol (TCP) and other compounds, at a fixed equilibrium concentration, as the initial concentration of the target compound in the isotherm test decreased.
The PAC residence time in an adsorption system must be long enough for the adsorption reactions to reach equilibrium if the capacity predicted by an equilibrium model is to be achieved. Although PAC has higher adsorption rates than granular activated carbon, it may still take several hours or more to reach equilibrium. For example, Najm et al. (1990b) found that eight hours was not sufficient time for TCP (initial concentration = 34 i~g/L, carbon dose = 5 mg/L) to come to equilibrium with the WPH PAC
(98% < 44 Ixm; geometric mean diameter = 10 ~zm). Because PAC is simply added to existing conventional treatment processes in many instances, and the residence time provided is usually in the range of one to two hours or less, an equilibrium model may greatly underestimate the
203
J. Environ. Eng., 1994, 120(1): 202-218

carbon dose required. In these instances, the use of an adsorption mass transfer model should be considered.

Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 10/23/15. Copyright ASCE. For personal use only; all rights reserved.

Mass Transfer Models

The homogeneous surface diffusion model (HSDM) is commonly used to describe adsorption kinetics. The model assumes that instantaneous equilibrium between the carbon particle and the adsorbate occurs at the outer surface of the carbon particle, and that the adsorbate moves to vacant sites inside the particle by surface diffusion. Typically, the HSDM requires input data such as the Freundlich K and 1/n for the adsorbate, and a set of batch kinetic data to determine the liquid film mass transfer coefficient, ks, and the surface diffusion coefficient, D~. These parameters are obtained by minimizing the difference between the concentration versus time data predicted by the HSDM and the measured batch kinetic test data. Different minimization procedures have been developed and used that intuitively vary the kinetic parameters until the best fit of the HSDM output to the batch data is obtained (Crittenden and Weber 1978a; Thacker 1981; Hand et al.
1983; Traegner and Suidan 1989).
Several studies have adequately demonstrated the use of the HSDM for describing the adsorption from one-adsorbate and two-adsorbate solutions onto activated carbon in batch and column contactors (Thacker 1981; Crittenden and Weber 1978b; Lee et al. 1983; Traegner and Suidan 1991). Some researchers (e.g., Najm et al. (1990a)] have also applied a pseudo singlesolute HSDM to predict the adsorption of a target compound in the presence of background organic matter. This approach assumes that the relation between the liquid-phase and solid-phase concentrations at the outer surface of the carbon particle is given by the isotherm of the target compound with an initial concentration equal to the initial concentration in the reactor, or to the influent concentration to the reactor. The adsorption of the target compound from the water is then treated as a single-adsorbate system.
Most studies of adsorption kinetics have used granular activated carbon
(GAC) as the adsorbent, but a few have used PAC. The primary difference in adsorption systems using these materials is that GAC is usuaUy applied using fixed beds, while PAC is applied as a slurry and moves with water being treated until it is removed by sedimentation and/or filtration. PAC can also be applied to water using fluidized beds (Lettinga et al. 1978; Najm et al. 1991a). Kim and Pingel (1989) developed a kinetic model to describe the removal of toluene from air by a PAC/water CSTR. The model incorporated absorption of toluene by the water, followed by adsorption of the toluene from water by the PAC. They used a single-solute mass transfer model that included both film and intraparticle diffusion as rate limiting steps, and solved the model using the technique of orthogonal collocation.
Good agreement between model predictions and experimental data was observed for a bench-scale reactor under both steady-state and unsteadystate conditions. Intraparticle diffusion was found to be the primary ratecontrolling step.
Traegner and Suidan (1991) developed an adsorption kinetic model for
PAC in a steady-state CSTR using the assumption that the surface diffusion was the sole rate limiting step. For single-solute adsorption, an analytical solution (Nakhla et al. 1989) was derived following the work of Crank
(1964). This analytical solution was used by Najm et al. (1990b) to describe adsorption of TCP from ground water by PAC in a bench-scale CSTR.
Although the effect of background organic matter in the ground water was
204

J. Environ. Eng., 1994, 120(1): 202-218

Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 10/23/15. Copyright ASCE. For personal use only; all rights reserved.

considered using the EBC method (Najm et al. 1991b), they were not able to obtain good predictions. Adham et al. (1991) later showed that the agreement between prediction and measured performance was affected by a slow reaction between TCP and dissolved oxygen on the carbon surface.
Adham et al. (1991) used a modified approach that involved searching the batch kinetic data to find the surface diffusivity, D~, using an apparent isotherm capacity parameter, Ka, determined from the batch-test data rather than the isotherm. The 1/n value was assumed to be the same as for the
TCP adsorption isotherm in ground water with the same initial concentration used in the batch test. The kinetic and equilibrium parameters thus obtained gave accurate predictions of the performance of a pilot-scale PAC/ultrafiltration system for removal of TCP. However, the drawback to this modified approach, is that the evaluated Ds and Ka values will unlikely be adequate to predict reactor performance at concentrations other than the one used in the batch test.
OBJECTIVE AND APPROACH

The objective of this research was to develop a procedure for evaluating adsorption of synthetic organic chemicals (SOCs) on PAC using the pseudo single-solute HSDM and independent determination of isotherm constants.
The procedure should yield accurate predictions of the adsorption kinetics of a SOC from waters containing background organic matter for a wide range of initial SOC concentrations.
The approach that was followed included (1) Writing the HSDM such that intraparticle diffusion was the only rate-limiting step, because Najm et al. (1990a) found that film diffusion was not important for PAC; (2) determination of the EBC parameters of the background organics using the isotherms of the target compound in distilled water and in natural water;
(3) incorporation of a procedure for calculating the Freundlich isotherm K value by the EBC method into the model solution procedure; (4) determination of the diffusion coefficient by applying the HSDM to batch kinetic data; (5) prediction of the performance of both the batch reactor and the
CSTR at different initial/influent concentrations of target compound and carbon doses; and (6) determination of the accuracy of the predictions by comparing them with the performance of a laboratory-scale batch reactor and two CSTRs, one of which was a PAC/ultrafiltration system.
MODEL DEVELOPMENT

HSDM for PAC in Batch Reactor

The adsorption of a target compound from natural water in a batch reactor was modeled by the pseudo single-solute HSDM. The model presented by
Crittenden and Weber (1978a) and others was rewritten such that intraparticle surface diffusion was the only rate-limiting step. This yields the following system of equations:
Ot

r 2 Or

r2

t=O,

O-0,

r = R:

Cs = C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

t = 0:

C = C1,o . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

(4)
(5)
(6)

where q = the solid-phase concentration of the compound; D~ = the surface diffusion coefficient; r = the radial distance from the center of the particle; t = the time; R = the PAC particle radius; Cl,o and C = the liquid-phase concentrations of the target compound in the bulk initially and at time t, respectively; C~ = the liquid-phase concentration of the target compound at the external surface of the PAC particle; and Cc = the PAC dose. Eq.
(1) is the HSDM equation that describes the mass transport of the target compound within a spherical carbon particle. Eq. (2) assumes that no adsorbate is present in the particle at the initial time. Eq. (3) assumes that there is no flux across the center of the particle. Eq. (4) states that the concentration change with time in the bulk solution is always equal to the change with time of the total average solid phase concentration.
To solve the aforementioned system of equations, the relationship between the liquid-phase concentration (Cs) and the solid-phase concentration
(qs) at the external surface of the particle must be provided. The Freundlich equation was used for this purpose: q~ : K(Cx.0, Cc)C~ 'n

9........................................

(7)

where 1/n is assumed to be equal to the single-solute Freundlich exponential constant for the target compound in distilled water; and K(CI,o, Co) = the capacity parameter, which is a function of the carbon dose (Co) and the initial concentration (C1.0) of the target compound in a specific natural water.
K(CI.o, C~) is determined by: (1) Finding the EBC parameters of the natural water for the target compound using the method of Najm et al. (1991b);
(2) determining the solid-phase and liquid-phase concentrations of the target compound at equilibrium by the IAST using the carbon dose, the initial concentration, the distilled-water isotherm parameters of the target compound, and the EBC parameters of the background organic matter as input data; and (3) calculating K(CI.o, Co) using (7) and the calculated solid-phase and liquid-phase concentrations.
The aforementioned system of pseudo single-solute HSDM equations is identical in form to the system of single-solute HSDM equations for the same reactor. Any numerical solution procedure that can be used to solve the single-solute HSDM equations can be used to solve the pseudo singlesolute HSDM. The only requirement is that the equilibrium parameter,
K(C~,0, Co), must be provided for each batch-test condition. In this study, the HSDM was solved numerically by a computer program using the orthogonal collocation technique (Finlayson 1972; Traegner and Suidan 1989).
The kinetic parameter, D~, was determined by analyzing experimental batch kinetic data with a search program that was based on the Levenberg-Marquardt unconstrained optimization algorithm and a numerical solution of the HSDM. Both programs were written by Traegner and Suidan (1989).
HSDM for PAC in CSTR

Application of PAC in basins such as the rapid mix or flocculation basin of a water treatment plant can be modeled as a steady-state CSTR. In such a reactor, aqueous adsorbate concentration is assumed to be constant and
206

J. Environ. Eng., 1994, 120(1): 202-218

Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 10/23/15. Copyright ASCE. For personal use only; all rights reserved.

equal to the effluent concentration, and the residence time of PAC particles follows an exponential distribution. Incorporation of the pseudo singlesolute HSDM assumptions described previously into the mass balance equation for the target compound leads to (8),

f i n - Ceff
-

CcK(Cin, t~f'x/n[1-6~l (
"~cj'-'eff

~-~ "= i 2 1 +

lsielr2,r~] = 0 9 (8)
R2 //j

This equation is identical to the equation first developed by Nakhla et al.
(1989) except that K(Cin , Cr appears in place of the single-solute Freundlich
K-value. In the equation, Cin and Ce, are the influent and the effluent concentrations of the target compound, respectively, K(Ci,, Cc) is the capacity parameter, which is a function of the influent concentration (Ci,) and the carbon dose (Cc) and can be calculated using the procedure described previously, and r is the average PAC residence time. A computer program was written to solve the model using the NEQNJ subroutine in Math (1990).
EXPERIMENTAL MATERIALS AND METHODS

Materials

The compound selected for testing the method was C TMatrazine, provided by Ciba-Geigy Corp., Greensboro, N.C. It had a specific activity of 53 t~Ci/ mg, a radiochemical purity of 94.1%, and a chemical purity of 96.5%.
Atrazine has a molecular weight of 215.7 g/mole.
The water sources used were deionized-distilled water and Central Illinois ground water. The deionized-distilled water made in the laboratory contained total organic carbon (TOC) between 0.1 to 0.3 mg/L. The ground water was pretreated using a green-sand filter to remove iron and manganese, and had the following characteristics: a pH of 7.3, 3 - 4 mg/L of
TOC, 307 mg/L of total hardness as CaCO3, and 290 mg/L of total alkalinity as CaCO3 (Najm et al. 1990b).
A stock solution was prepared by spiking atrazine into water without the use of nonaqueous solvents. The spiked solution was stirred for about two days in the dark. Because the crystalline atrazine was found to be not fully dissolvable, the stock solution was filtered through 0.45-1xm membrane filter paper before it was stored at 4~ in a dark room for later use. For the deionized-distilled water solution, a 0.001-M phosphate buffer was added to maintain the pH at about 7.3; buffer was not needed for the ground water since it contained enough alkalinity to maintain a constant pH. All working solutions for the experiments were prepared from the stock solution.
The PAC used was WPH powdered activated carbon (Calgon Carbon
Corp., Pittsburgh, Pa.). It had a geometric mean diameter of about 10 Ixm
(determined using an Elzon Particle Counter, Elmhurst, Ill.). The PAC was randomly taken from a bulk bag, dried at 105~ for a few hours, and stored in an air-tight desiccator prior to use.
Isotherm Test

The adsorption capacity data at equilibrium were gathered by contacting chemical solutions of known initial concentrations with different quantities of PAC for a period of time (Randtke and Snoeyink 1983). Measured quantities of PAC were placed into several 250-, 500-, or 1,000-mL brown
207

J. Environ. Eng., 1994, 120(1): 202-218

Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 10/23/15. Copyright ASCE. For personal use only; all rights reserved.

glass bottles to which known amounts of solution were added. Controls (at least three bottles) that contained the same solution but no carbon were part of each experiment. The bottles were placed on a mechanical shaker for seven days. The solutions were separated from the PAC by filtering through prewashed 0.45-1xm nylon membrane filters (Magna, Westboro,
Mass.).
Batch Kinetic Test

The apparatus for batch kinetic experiments consisted of a 4-L glass jar and an overhead stirrer with a speed controller. To conduct a batch experiment, a measured volume of the stock solution was added to a known volume of water in the jar and then the solution was stirred for an hour to ensure complete mixing. Prior to the addition of a measured quantity of
PAC, triplicate samples were taken to determine the initial concentration of atrazine. The PAC was soaked in 10-mL deionized-distilled water for about four hours before addition. A variable-speed motor was used to maintain complete mixing throughout the batch-test run. After adding the
PAC, 2.5-mL samples were withdrawn periodically using 2.5-mL luer-lock, gas-tight glass syringes (Hamilton Co., Reno, Nev.) and were immediately filtered through prewashed 0.45-p~m nylon membrane filters (Magna, Westboro, Mass.).
Continuous Stirred Tank Reactor

A 5.25-L CSTR that was stirred with a magnetic stirrer was used in this study (Fig. 1). Atrazine solution and PAC slurry were fed directly into the reactor by two positive-displacement Masterflex pumps (Cole-Parmer Instruments Co,, Chicago). In each CSTR experiment, the reactor was fully filled (no headspace) with the PAC-atrazine solution. Because the reactor was covered, the solution flowed out of the reactor under its own pressure.
The reactor was run for three mean retention times before the effluent samples were taken and analyzed. The samples were taken using 2.5-mL luer-lock, gas-tight glass syringes (Hamilton Co., Reno, Nev.), and were

EIqT

PAC FEED
~OLUTIOH
FIG. 1. Schematic Diagram of CSTR Experimental Setup
208
J. Environ. Eng., 1994, 120(1): 202-218

Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 10/23/15. Copyright ASCE. For personal use only; all rights reserved.

immediately filtered through prewashed 0.45-~m nylon membrane filters
(Magna, Westboro, Mass.),
PAC/UF Pilot Unit
A schematic diagram of the PAC/ultrafiltration (PAC/UF) pilot system
(Nautilus, Lyonnaise des Eaux-Dumez, Le Pecq, France) is provided in Fig.
2. The feed-mix tank and the recycle loop could be analyzed as a CSTR.
The positive displacement pump provided the necessary pressure and recirculation. The pulsation absorber dampened pulsations from the pump.
The pilot unit was operated at constant permeate flow rate with variable pressure. This operational mode allowed better control over both the mean hydraulic and PAC residence times. Mean hydraulic residence time was determined by dividing the volume of water in the recycle loop (which includes the mixing tank and tubing) by the influent flow to the mixing reactor. The PAC was removed from the system by continuously wasting water from the mixing reactor at a preset flow rate. Variation of the PAC mean residence time was achieved by changing the PAC purge flow rate.
The PAC mean residence time was determined by dividing the volume of the recycle loop by the purge flow rate; the PAC concentration was assumed to be uniform throughout the recycle loop. The pilot unit was operated for a period of at least three PAC mean residence times before steady state was assumed and samples were collected from the influent (before blending with PAC) and permeate line for analyses.
The PAC/UF system was operated continuously in a semiautomatic mode.
Permeate flow rate, temperature, and hollow-fiber inlet/outlet pressures were periodically recorded. As the permeate flow rate was reduced due to fouling, the pressure was manually increased to maintain the preset constant permeate flow rate. When the inlet pressure reached 200 kPa (29 psi), backflushing was performed.
Atrazine Concentration Measurement
The atrazine used in this study was labeled with C TM. To measure its concentration, 2.5-mL sample was first mixed with 18-mL universal LSC

PO.ITIVE
DImPLACEMENT PUIV~

rLov
-

ATRAZINE ( ' ~
SOLUTION~ J

f ' ~ PAC
GROUNDVATER
_ . _ _ [ ~ ,LI/RRY ,

:::::::::::::::::::::::::::::::::::::
============================
FEED/MIX TANK

-

TO

UF CARTRIDGE

FIG. 2.

SEI~oR

~
BACEFLUSH NITROGEN
VESSEL
PRESSURE
TANK

Schematic Diagram of PAC/UF Pilot System
209

J. Environ. Eng., 1994, 120(1): 202-218

Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 10/23/15. Copyright ASCE. For personal use only; all rights reserved.

cocktail (SX16-4, Fisher Scientific) by hand-shaking for one minute. Then
C t4 was counted for 20 minutes with a liquid scintillation analyzer (Packard
Instrument Co., Downers Grove, Ill.).
RESULTS AND ANALYSIS

Batch Equilibrium Study

Fig. 3 shows the adsorption isotherms of atrazine in deionized-distilled water and ground water. The deionized-distilled-water Freundlich isotherm parameters, 1/n and K, were 0.335 and 797 (limole/g) (L/p~mole) TM, respectively. These parameters are comparable to the parameters determined by Speth and Miltner (1990) for adsorption of atrazine on the pulverized
F-400 granular activated carbon (1/n = 0.291 and K = 858 (p~mole/g) (L/
;xmole)l/").
The two ground-water isotherms had different atrazine initial concentrations. As shown in Fig. 3, the initial concentration had a significant impact on the isotherm in the ground water. The lower the initial concentration used to conduct the isotherm, the lower the surface loading of atrazine was at the same liquid-phase concentration. This behavior was apparently caused by the competition from the background organic matter in the water. Application of the EBC method to the isotherm with the initial concentration of 41.5 p,g/L led to the EBC parameters of C2,0 = 2.2 p,mole/L, 1/n2 = 0.42 and/(2 = 650 (limole/g) (L/p,mole) TM for the background organic matter.
With these EBC parameters and the single-solute isotherm parameters for atrazine, the lAST model adequately fit the ground-water isotherm for atrazine with an initial concentration of 41.5 Ixg/L and successfully predicted the ground-water isotherm for atrazine with an initial concentration of 215 p~g/L as shown in Fig. 3. The predicted isotherm for an initial concentration of atrazine of 4.5 ~g/L in ground water is also shown in Fig. 3.
1000

9

In Dcionizcd-Disti11rWater

o Cl,o= 215 l.tgrb Groundwater in 9 CI,o= 41.5 lig/Lin Groundwater
:

10o

. -"

lAST P e c on,

g

.....-a'~

~ ''-

in C_n'oundwatcr
1

.I

1

I0

100

1000

Liquid Phase Cone., gg/L
FIG. 3. Atrazine Isotherms in Deionized-Distilled Water and in Ground Water with lAST Predictions

210
J. Environ. Eng., 1994, 120(1): 202-218

Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 10/23/15. Copyright ASCE. For personal use only; all rights reserved.

Batch Kinetic Study
Fig 4 shows the batch kinetic test data for adsorption of atrazine from deionized-distilled water and the HSDM fit of these data. Adsorption equilibrium was not reached within five hours The surface diffusion coefficient that best fit the experimental data was 7 9 • 10 -11 cm2/min
Fig 5 shows two sets of batch kinetic test data for adsorption of atrazine

1.2
L
1.01 o Atrazine C 1.o = 252 ~g/L
PAC Dose = 2.6 mg/L

0.8

[]

HSDMFit

[]

0.6
0.4
0.2
0.0 /

,

,

0

,

50

,

,

,

100

,

.

] 50

,,

,

200

,

250

300

Time, rain.
FIG. 4. Batch Kinetic Test Results for Atrazine in Deionized-Distilled Water and
HSDM Fit

,.2i o Atrazine C1,0 = 22.5 lxg/L
PAC Dose -- 3.3 mg/L

1.0

Data (fromtest #1) o Data(fromtest #2)

0.8
0

0.6
<

0.4
0.2 i
0.0
0

I

50

I

I

9

I

,

I

,

I

~

I

,

I

,

100 150 200 250 300 350 400 450
Time, win.

FIG. 5,

Batch Kinetic Test Results for Atrazine in Ground Water and HSDM Fit
211

J. Environ. Eng., 1994, 120(1): 202-218

Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 10/23/15. Copyright ASCE. For personal use only; all rights reserved.

from ground water, both with the same initial concentration (C1,0), 22.5 txg/L, and the same carbon dose (Co), 3.3 mg/L. Good agreement between the two sets of data indicates that the experiment was reproducible. The capacity parameter, K(C1.0, Co), determined using the procedure described previously was 144.5 (p~mole/g) (L/limole) TM. This parameter along with the atrazine single-solute Freundlich 1In value were then used as input data in the pseudo single-solute HSDM for a batch reactor to determine the surface diffusion coefficient, D~. The ground-water D~ value determined for the best fit of both sets of the experimental data was 3.7 x 10 T M cruZ/rain, approximately one-half the value determined for deionized-distilled-water system.
In order to test whether the ground-water D~ value determined was valid under other conditions, model predictions were made for different carbon dosages and initial concentrations. Experiments were then conducted using these conditions to determine the accuracy of the predictions. Fig. 6 shows the data for initial concentrations of 4.5 and 222 Ixg/L, and the corresponding model predictions. Excellent agreement was observed indicating that the
Ds value determined was unique and not a function of initial concentration or carbon dose over the wide range tested.
The capacity parameter, K(CLo, Cc), decreased exponentially as the initial concentration of atrazine decreased for a given PAC dose. For example, for a PAC dose of 2.5 mg/L, its value decreased from 91 (ixmole/g) (L/ ixmole) TM at C~,o = 10 lxg/L to 20.1 (ixmole/g) (L/ixmole) TM at C1,0 = 1.0 lxg/L. For a given initial concentration, K(C1,o, Co) decreased as the carbon dose increased. For example, at an initial concentration of 5 ixg/L, the
K-value decreased from 88.7 to 48.2 (txmole/g) (L/p,mole) TM as the PAC dose increased from 0.2 to 10 mg/L. This result implies that the atrazine isotherm at a given initial concentration in the ground water cannot be represented accurately by one Freundlich isotherm equation. This observation is consistent with the curved lines representing the atrazine isotherms

o1

e,

~
L

'"

s

"

Alrazine C 1,0= 4.5 ll~l~ & PAC Dose = 2.4 mg/L

1.0~ o Atrazine 222IJ.g/L PACDose= 12.9
Cl,o=
& rng/L 0"9k
0.8

' PredictedPerformance

03
0.5
<

0,4
0.3
0

.

O.l
0.0

I

0

I,

50

2
9

I

100

i

~
?

i.

150

! I

~

200

Bi

250

.

ml

300

.

'

350

Time, rain.
FIG. 6. diction Batch Kinetic Test Results for Atrazine in Ground Water and HSDM Pre-

212
J. Environ. Eng., 1994, 120(1): 202-218

in ground water that are shown in Fig. 3. It is also consistent with a decrease in PAC surface coverage with background organic matter as carbon dose increased. Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 10/23/15. Copyright ASCE. For personal use only; all rights reserved.

Predicting Performance of CSTRs

The experimental data collected from a bench-scale CSTR and a pilotscale PAC/UF system treating Illinois ground water were used to determine the validity of the CSTR model (8). The performance predictions used kinetic and equilibrium parameters that were obtained from independent batch tests using the same water. Predictions were made for initial concentrations and PAC doses other than those used in the batch tests.
Fig. 7 shows the results of a CSTR experiment in comparison to the predicted values. The CSTR had an average PAC residence time (mass
PAC in CSTRJPAC feed rate) of 54 minutes. A constant influent atrazine concentration of 17.4 p~g/L was treated with different carbon doses. The good agreement between the predicted performance and the experimental data shows that the procedure developed can be used with good reliability to describe adsorption of atrazine from ground water at this concentration.
Because the mixing tank and the recycle loop portion of the PAC/UF process can be considered to be a CSTR (Adham et al. 1991), the CSTR model was used to predict the performance of a PAC/UF pilot unit for atrazine removal. The pilot system was operated with a constant carbon dose but at different PAC residence times by changing the PAC wastage flow rate. (PAC residence time = mass PAC in reactor/mass PAC wasted/ minute.) Good agreement again was obtained between the CSTR model prediction and the experimental performance of the PAC/UF system as shown in Fig. 8.
The performance of a PAC system for drinking-water treatment is mainly determined by the following parameters: the influent concentration of the target compound, the contact time of the system, the type and the dosage

1.0 i

PAC RT = $4 rain.

0.8

Atrazine Cin = 17.4P4g/L
Ds = 3.7x10 cm2/min.
"11
9

6

Data
Prediction

0.6
0.4
0.2
0.0

F
--,

0

.

.

!

2

9

9

9

I

4

9

,

,

I

6

9

9

i

I

8

,

9

9

I

9

10

9

9

I

9

12

m

|

I

14

i

9

9

I

J

16

"
9

9

18

PAC Dose, mg/L
FIG. 7. CSTR Model Prediction versus Performance for Adsorption of Atrazine from Ground Water

213
J. Environ. Eng., 1994, 120(1): 202-218

AtrazineCin= 26.6
PACDoseCc = S.8 enm..,/L

1.I
~

Ds = 3.7x10-11 cm2/min.

Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 10/23/15. Copyright ASCE. For personal use only; all rights reserved.

0.8

\

~

0.4

<

PAC/UFExi~.ri:ental Data

0.6

.~

"

0.2

.0
0.0

I

1.0

~

I

I

I

I

I

2.0

3.0

4.0

5.0

6.0

7.0

PAC Residence Time, hr
FIG. 8. CSTR Model Prediction versus PAC/UF Performance for Adsorption of
Atrazine from Ground Water

of PAC, the effluent concentration required, and the influence of the background organic matter. For a given process design, one type of PAC, and one type and concentration of background organic matter, the contact time of the reactor is fixed, and the effluent concentration is determined by the
PAC dose. The dose will b e fixed to achieve a process goal, such as the maximum contaminant level (MCL) of 3 ~g/L for atrazine in the United
States. The only parameter that must be considered as an independent variable is the influent concentration of atrazine, since it may vary from day to day. Fig. 9 shows the PAC dose to achieve 3 I~g/L of atrazine in a CSTR with different PAC residence times as a function of influent concentration.
The background organic matter for this simulation is that in the ground water used in the experimental phase of this study. The one-hour contact time is thought to be typical for PAC addition to the rapid mix of a conventional treatment process, and the 24-hour contact time is typical for a
PAC floc-blanket reactor. As shown in Fig. 9, a carbon dose of 7.8 mg/L can reduce 12 ixg/L of atrazine in a CSTR influent to 3 Ixg/L in the effluent with a PAC residence time of one hour. However, if a longer contact time of 24 hours is provided, only 2.1 mg/L of PAC will be needed to treat the same water. This indicates that efficient use of PAC can be achieved using an adsorption process that would allow a long contact time, such as PAC addition to a floc-blanket reactor or the PAC/UF process.
In Europe, the standard for atrazine is 0.1 txg/L (Carney 1991). Fig. 10 shows the PAC doses required to achieve the 0.1 Ixg/L MCL of atrazine from Illinois ground water using a CSTR with different contact times. In a contactor with one-hour PAC residence time, 33 mg/L of PAC is needed to reduce 12 Ixg/L of atrazine to 0.1 Ixg/L. In a contactor with a 24-hour residence time, however, only 10.7 mg/L of PAC is needed to treat the same water. This shows the significant impact of the regulation on the PAC
214

J. Environ. Eng., 1994, 120(1): 202-218

8
7
Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 10/23/15. Copyright ASCE. For personal use only; all rights reserved.

6

PAC RT = I hr

5 o 4

o

3
2
1
0

3

2

4

5

6

7

8

9

10

11

12

Influent Conc., gg/L
FIG. 9. Carbon Dose Needed for Atrazine Effluent Concentration of 3 i~g/L from
CSTR as Function of Influent Concentration and PAC Residence Time

35

25

=

=

20
15

~

~

~

10

~"

i

0

0

I

1

i

|

2

i

|

3

.

I

4

i

I

5

s

I

6

i

I

|

I

7

8

9

I

I

10 11 12

Influent Conc., I.tg/L
FIG. 10. Carbon Dose Needed for Atrazine Effluent Concentration of 0.1 i~g/L from
CSTR as Function of Influent Concentration and PAC Residence Time

usage rate. Figs. 9 and 10 indicate that the carbon usage rate is more sensitive to variation of the influent concentration if the PAC contact time is short.
Hence, a PAC adsorption system with a longer contact time would have a higher resistance to sudden influent concentration variation.
215
J. Environ. Eng., 1994, 120(1): 202-218

Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 10/23/15. Copyright ASCE. For personal use only; all rights reserved.

SUMMARY AND CONCLUSIONS

A procedure was developed for predicting the removal of trace organic pollutants from natural water by PAC adsorption systems. The method uses the EBC method coupled with the lAST to quantify the competition of the background organic matter, and uses the pseudo single-solute HSDM to describe the adsorption kinetics of the target compound. With the parameters obtained from independent batch equilibrium and kinetic tests, the performance of a continuous-flow PAC adsorption system can be predicted.
The continuous-flow PAC adsorption system was assumed to be a CSTR, which we believe is a reasonable representation of the rapid mixing and flocculation basins in a conventional water treatment plant. The method was applied and verified for the adsorption of atrazine from Central Illinois ground water using a bench-scale CSTR and a PAC/UF pilot unit. The following conclusions were made:
1. The PAC adsorption capacity for atrazine was reduced due to competitive adsorption of the background organic matter. This capacity reduction was accurately quantified by the EBC method of Najm et al. (1991b) coupled with the IAST.
2. The pseudo single-solute HSDM satisfactorily described the adsorption kinetics of atrazine under the influence of the background organic matter.
The experimentally determined surface diffusion coefficient, Ds, was unique and not a function of initial concentration or carbon dose over the wide range tested. Comparison of Ds values for the adsorption of atrazine from deionized-distilled water and from the ground water indicated that the intraparticle mass transfer rate for atrazine was reduced by a factor of about two due to the adsorption of the background organic matter.
3. Good agreement between the CSTR model predictions and the experimental data from the bench-scale CSTR and the PAC/UF pilot unit was observed. 4. The continuous-flow PAC adsorption systems with longer PAC contact times were more efficient for the use of PAC to remove atrazine. Systems with longer PAC contact times are also less sensitive to sudden influent concentration variation.
ACKNOWLEDGMENTS

Funding for this work was provided by the U.S. Environmental Protection
Agency, grant No. CR 814034, the American Water Works Association
Research Foundation, and Lyonnaise des Eaux-Dumez, France. The PAC was provided by Calgon Carbon Corp., and Ciba-Geigy provided the radiolabeled atrazine. This paper has not been subjected to peer review by the sponsors; the opinions expressed are those of the writers and do not necessarily reflect the views of the sponsors. Official endorsement of the sponsoring agencies should not be inferred.
APPENDIXI.

REFERENCES

Adham, S. S., Snoeyink, V. L., Clark, M. M., and Bersillon, J.-L. (1991). "Predicting and verifying organics removal by PAC in an ultrafiltration system." J. A m . Water
Works Assoc., 83(12), 81-91.
Carney, M. (1991). "European drinking water standards." J. A m . Water Works
Assoc., 83(6), 48-55.
216
J. Environ. Eng., 1994, 120(1): 202-218

Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 10/23/15. Copyright ASCE. For personal use only; all rights reserved.

Clark, S. W. (1989). "Overview of regulatory impacts in GAC use." Proc. of a
Technology Transfer Conf. on Design and Use of Granular Activated CarbonPractical Aspects, American Water Works Association Research Foundation, Cincinnati, OH.
Crank, J. (1964). The mathematics of diffusion. Oxford Univ. Press, London, England.
Crittenden, J. C., and Weber, W. J. Jr. (1978a). "Predictive model for design of fixed-bed adsorbers: Parameter estimation and model development." J. Envir.
Engrg. Div., ASCE, 104(2), 185-197.
Crittenden, J. C., and Weber, W. J. Jr. (1978b). "Predictive model for design of fixed-bed adsorbers: Single-component model verification." J. Envir. Engrg. Div.,
ASCE, 104(3), 433-443.
Finlayson, B. A. (1972). Method of weighted residuals and variational principles.
Academic Press, New York, N.Y.
Hand, D. W., Crittenden, J. C., and Thacker, W. E. (1983). "User-oriented batch reactor solutions to the homogeneous surface diffusion model." J. Envir. Engrg.
Div., ASCE, 109(0, 82-101.
Kim, B. R., and Pingel, L. J. (1989). "Removal of toluene from air using PAC/ water slurry reactor." J. Envir. Engrg. Div., ASCE, 115(5), 1025-1045.
Lee, M. C., Crittenden, J. C., Snoeyink, V. L., and Ari, M. (1983). "Design of carbon beds to remove humic substances." J. Envir. Engrg. Div., ASCE, 109(3),
631-645.
Lettinga, G., Beverloo, W. A., and Vanlier, W. C. (1978). "The use of flocculated powdered activated carbon in water treatment." Progress in Water Tech., 10(1/2),
537-554.
Math~PC-Library. (1990). International Mathematics and Statistical Library (IMSL),
Houston, Texas.
Miltner, R. J., Baker, D. B., Speth, T. F., and Fronk, C, A. (1989). "Treatment of seasonal pesticides in surface water." J. Am. Water Works Assoc., 81(1), 4352.
Najm, I. N., Snoeyink, V. L., Suidan, M. T., Lee, C. H., and Richard, Y. (1990a).
"Effect of particle size and background natural organics on the adsorption efficiency of PAC." J. Am. Water Works Assoc., 82(1), 65-76.
Najm, I. N., Snoeyink, V. L., Galvin, T. L., and Richard, Y. (1990b). Evaluation of powdered activated carbon use for the control of organic compounds during drinking water treatment. American Water Works Association Res. Foundation,
Denver, Colo.
Najm, I. N., Snoeyink, V. L., Lykins, B. W., and Adams, J. Q. (1991a). "Powdered activated carbon for drinking water treatment: A critical review." J. Am. Water
Works Assoc., 83(1), 65-76.
Najm, I. N., Snoeyink, V. L., and Richard, Y. (1991b). "Effect of initial concentration on its activated carbon adsorption capacity in natural water." J. Am. Water
Works Assoc., 83(8), 57-63.
Nakhla, G. F., Suidan, M. T., and Traegner-Duhr, U. K. (1989). "Steady state model for the expanded-bed anaerobic GAC reactor operating with GAC replacement and treating inhibitory wastewaters." Proc. Industrial Waste Symp., 62rid
WPCF Annual Conf., WPCF, Washington, D.C.
Radke, C. J., and Prausnitz, J. M. (1972). "Thermodynamics of Multisolute Adsorption from Dilute Liquid Solutions." J. Am. Inst. Chem. Engrg., 18(4), 761768.
Randtke, S. J., and Snoeyink, V. L. (1983). "Evaluating GAC adsorption capacity."
J. Am. Water Works Assoc., 75(8), 406-413.
Sontheimer, H., Crittenden, J. C., and Summers, R. S. (1989). Activated carbon for water treatment. American Water Works Association Res. Foundation, Denver,
Colo.
Speth, T. F., and Miltner, R. J. (1980). "Technical note: An evaluation of GAC adsorption capacities for SOCs." J. Am. Water Works Assoc., 82(2), 72-75.
Thacker, W. E. (1981). "Modeling of activated carbon and coal gasification char
217

J. Environ. Eng., 1994, 120(1): 202-218

Downloaded from ascelibrary.org by TEXAS A&M UNIVERSITY on 10/23/15. Copyright ASCE. For personal use only; all rights reserved.

adsorbents in single-solute and bi-solute systems," PhD thesis, Univ. of Illinois,
Urbana, Ill.
Traegner, U. K., and Suidan, M. T. (1989). "Parameter evaluation for carbon adsorption." J. Envir. Engrg., ASCE, 115(1), 109-128.
Traegner, U. K., and Suidan, M. T. (1991). "Steady state performance of activated carbon contactors." Water Sci. Tech., 23(7-9), 1677-1686.
APPENDIX !1. NOTATION

The following symbols are used in this paper:
C = liquid-phase concentration of target compound in bulk solution at time t (M/L3);
C1 ,o = initial liquid-phase concentration of target compound in bulk solution (M/L3);
C2,0 = initial concentration of equivalent background compound
(M/L~);
c~ = carbon dose (M/L3);
= CSTR effluent concentration of target compound (M/L3); c,. -- CSTR influent concentration of target compound (M/L3); c, = liquid-phase concentration of target compound at external surface of PAC particle (M/L3);
D, = surface diffusion coefficient (L2/T);
K = Freundlich isotherm capacity of target compound in singlesolute system (M/M)(L3/M)I/";
K= = Freundlich isotherm capacity of equivalent background compound (M/M)(L3/M)Vn;
K(C1/o, Cc) = Freundlich capacity parameter, which is function of Cl,o and
Cc (M/M)(L3/M)I/n;
K(C~., Cc) = Freundlich capacity parameter, which is function of Ci, and
C~ (M/M)(L3/M)I/"; k~ = film mass transfer coefficient (L/T); n = Freundlich isotherm constant of target compound in singlesolute system (dimensionless); n2 = Freundlich isotherm constant of equivalent background compound (dimensionless); q = solid-phase concentration of target compound (M/M); q~ = solid-phase concentration of target compound at external surface of PAC particle (M/M);
R = carbon particle radius (L);
RT = residence time (T); r = radial distance inside PAC particle (L); t -- time of adsorption (T); and
T = average carbon residence time (T).

218
J. Environ. Eng., 1994, 120(1): 202-218