Prediction of Specific Priority Organics Adsoption on Activated Carbon in Complex Background Mixtures

A study was conducted to investigate carbon adsorption of specific priority pollutants, (2,4 Dimethylphenol, Naphthalene, Fluorene, Pyrene, Bis(2-ethylhexyl phthalate) in a complex mixture background, using laboratory and pilot scale reactors for data generation , and various numerical computer models for data evaluation and subsequent prediction. Prediction of adsorption using intra-particle diffusion resistance (surface model and combined surface and pore diffusion model) was more successful using the combined surface and pore diffusion model than by using the surface model alone . Dissolved organic carbon (DOC) was selected as a complex mixture surrogate to evaluate any competitive effect due to individual components in the complex background mixture. Prediction of batch and packed bed adsorption of each of the specific priority organics that were studied was successful when the IAS model was adjusted by evaluating the adsorption equilibrium constants in a complex mixture background. Results from the laboratory scale experiments were verified by a pilot plant study utilizing 4 inch diameter by 3 feet deep carbon contactors with empty bed contact time of 1.75 and 3,75 minutes treating 2,4-dimethylphenol present in a complex mixture background consisting of a domestic secondary treatment effluent. ACKNOWLEDGEMENTS This project was funded by the Kuwait Institute for Scientific Research and the Kuwait Environmental Protection Council to whom the author feels deeply indebted. Without the awareness and concern of their directors and staff, this project would not have been made possible. I wish to express my sincer e gratitude to my advisor Dr. Leon T. Thiem for all his support, guidance and supervision throughout this project. I would also like to thank my committee members Dr. Calvin P.C. Poon for his valuable comments and suggestions, Dr. Raymond Wright for his valuable comments, suggestions and all those hours of side meetings we had, and to Dr. William Wright for his r eview and comments. Gratitude is extended to Milton and George Huston for all their help in the various stages of project progress as well as in constructing the pilot plant. Their skills and dedication permitted the timely completion of this study. Appreciation is sincerely expressed to the staff of South Kingstown Wastewater Treatment Facility for their cooperation and help in installing the pilot plant in a safe and secure lucation. My appreciation to Mary Costa and Kevin Fish for their excellent typing and patience in putting up with the quality of my handwriting. Unend i ng support, patience and encouragement from my wife Fairouz is greatly acknowledged.

All of the wastewater that is generated in SIA (about 30,000 m3/d) is discharged back to the sea with minor in-plant treatment .
The available in-plant treatment includes processes such as NH3 stripping, flue gas desulfurization units, urea hydr olizers, API oil separators and neutralization pits. The main water-wastewater cycle in SIA is shown in Figure 1 .1. Treatment of SIA wastewater is a critical issue due to the following.
The desalination plant water intakes are located in the vicinity of the wastewater o utfalls. Several low vapor pressure pollutants that are discharged from the SIA industries to the off-shore waters may carried over during the distillation process that is used to produce drinking water for Kuwait. This is of concern since this distilled water constitutes the major source of drinking water for The major limitation to much needed agriculture in Kuwait is the lack of sufficient and suitable water for irrigation.
This has led to the concensus that a non-discharge policy of wastewater needs to be adopted in this area.
With a program of wastewater reclamation and reuse, the treated industrial water can be recycled to the industries , utilized for non-crop irrigation, and can replenish brackish water aquifers. A previous study was conducted by the author1 in order to investigate the central treatability of all of the wastewater generated from SIA. The study involved the operation of a pilot plant used to model a combined carbon oxidationnitrification process, employing a continuous flow stirred-tank reactor with recycle and zero sludge wasting mode of operation.
The overall conclusion of the study indicated that the combined treatment of SIA industrial wastewater is feasible and adequate for removing conventional pollutants. However, one of the major concerns is the probable presence of organic priority pollutants which are generated by the petroleum industries and can impose substantial health hazards and restrictions on reuse alterna-tives2-6.
In order to provide the maximum use of the wastewater generated in SIA with minimum health concern restrictions, a polishing of the effluent derived from the activated sludge treatment system would be essential. This concept of adding treatment units to the conventional secondary treatment sequence provides the model for Best Available Treatment Economically Achievable (BATEA) for this kind of wastewater. Most promising polishing techniques include powder activated carbon (PAC) and granular activated carbon (GAC) processes. The application of this treatment sequence can provide a high quality reclaimed wastewater that can be used for a variety of applications.

.2 Wastewater Characteristics
A combined petrochemical industrial complex wastewater is expected to contain a variety of priority organic pollutants depending on the nature of the processes. Many of these pollutants are common among the different industrial categories and others are specific.
With the lack of information on petrochemical industries wastewater characteristics it is very difficult to establish a criterion for petrochemical wastewater.
Petroleum refining industrial wastewater is considered to have the highest levels of organic priority pollutants. Wastewater in the petroleum refining industry is defined as the effluent from an API Separator which is only designed to remove gross concentration of oil and grease.
For the purpose of establishing wastewater characteristics of the petroleum refining industry, literature data were reviewed and supplemented with actual sampling and analysis. The characteristics of petroleum refineries wastewater are presented in ------a : Adlpttd fr• raftl"tllCts (2,3) lna~s of 24-hr. c•posfte s•plts on 3 consecutht days) . II : Slluefba lndllltrfal Are1 eo.1ax (S fad Area). c : Analyzed for the purpose of this stuq (Avaraga of tltrtt 24-hr. c•posfte 1 .. plts) . d: Ad1Pted from reftl"tllCt (4) . NA : llot 1n1lyztd. Ill: Not cMtecttd. Rtfn .: Rllffntl")' samples were collected in 24 hour composites on three consecutive days and analyzed for the presence of the 129 priority pollutants, and other conventional parameters. Table 1-1 also incl udes the results of analysis performed on one of SIA major refineries (KNPC-Shuaiba). Three 24-hour composite samples of the API separator effluent were analyzed for priority pollutants over a one month period of time.

Selection of Representative Priority Organics
Based on the data presented in Table 1, the representative priority organics of refinery wastewater which were selected for use in this study were: 1. Bis (2-ethylhexyl)

Fluroene
The reasons for their selection are the following: 1. These compounds are potential health hazards and exhibit toxic, carcinogenic or mutagenic properties.
2. The selected compounds have been measured in most of the surveyed refineries including the SAA refineries.
3. Compounds selected were representative of the major chemical groups within the priority organics compounds typically found in refinery wastes. 4. Volatile priority pollutants were excluded since significant removals would be expected to occur by treatment processes preceeding GAC adsorption such as dissolved air flotation.

.4 Objectives of the Study
Most of the research previously conducted or activated carbon adsorption has focussed on the adsorption of a single solute or a multi-solute mixture of known composition. Within the past few years the adsorption of specific organics present in unknown background composition has begun to be studied. These studies only considered ground water which was contaminated with volatile organics. Compared to the types and levels of organic chemicals present in a refinery wastewater the groundwater background organic chemical composition is close to that of a pure water.
The performance of models developed for predicting carbon adsorption of individual solutes in the presence of a relatively contaminated water such as a refinery wastewater need to be examined. Accordingly the objectives of this research project are the following:

1.
Evaluate the adsorption isotherm parameters of the following compounds; 1) 2,4,dimethylphenol, 2) napthalene, 3) fluorene, 4) pyrene and 5) bis (2-ethylhexyl) phthalate. The isotherm parameters will be determined first as single solutes and then as multi-solutes in both an ultrapure water background and in a typical complex wastewater background.
Several isotherm models will be used for fitting the  (2)(3) where 6 U 0 = the difference in internal energy between the adsorbed and the dissolved adsor bate in solution  It is based on the assumption that maximum adsorption corresponds to a saturated monolayer of adsorbate on the adsorbent surface, all sites are energetically equivalent , and there is no interaction between adsorbate molecules on the neighboring sites.

Three Parameters Model
As an attempt to overcome the limited concentration application range of Langmuir and Freundlich models , Radke and Prausnit z 42 developed a three-parameter model that combines both Langmuir and Freundlich models . This model has the following relationship: Cl c e + a c Y e (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13) where, a, a. y can be estimated mathematically by fitting experimental data. As was intended originally by Radke

Thermodynamic
The fundamental thermodynamic equations representing the combined first and second laws may be written in four equivalent ways: 1) internal energy, 2) enthalpy, 3) Helmholtz free energy, and 4) Gibbs free energy. For an adsorbed phase the Helmholtz free energy in two dimensions designated by superscript "a" can be written as: (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17) where subscripts i and s represent solute and solvent respectively, o is the interfacial tension, µi is the chemical potential, ni is the number of moles adsorbed on the solid surface, A is the area of solution-solid interface , S is the entropy and T is the absolute temperature .
where the summation is over the solute species only. The spreading pressure 1f is defined as the difference between the inte r facial tension of pure solvent-solid interface and that of the solution-solid interface at same temperature. 1f = 0 pure solvent-solid-a solution-solid  The difference between qei as used previously and nim is that qei is for a specified adsorbent weight, while nim is for unspecified amount of adsorbent. The adsorbed-phase fugacity is two-dimensional and has the same units as the spr eading press ure n. Henr y 's Law for adsorption , results from the subs t itution o f equations ,  and  with Equation .

Ideal Ads orbed Phase
Ra dke & Prausnitz45 prop osed tha t when solute is adsorbed simultaneously from dilute solutions at constant tempera t ure and spreading pressure , then the fugaci t y fia is proportional to the mole f rac t ion zi f.a (T , n , Z. ) 1 1 Z.f.ao (T,n) 1 1 (2-29) where fia o is the fugacit y t hat a single solut e i exer ts during adsorption from dilute solu t ion at the same tem p era t ure and spreading pressure. The superscript "0" denotes single-solute adsorption. Equa t i on  becomes exact a s Zi~1. Substituting Equation  into equation  gives : A dTI  Rearranging and factoring out the total moles adsorbed, nTm, gives:  For an ideal adsorbed phase, equation , equation , and equation  gives:

Phase Equilibria
If a mixture of solutes is in equilibrium with an adsorbent, the chemical potential of solute i in the adsorbed phase is equal to that in the liquid phase "l"  The value of µia depends on T , TI, and the adsorbate composition as measured by the mole fraction Zi. To calculate the chemical potential, the integral form of equation  can be derived and substituted into the ideality assumption of equation (   where the supersc r ipt "#" denot e s the ideal di l ute so l ution (Henry's Law) standard state of the 1 i quid phase. The activity coefficient is considered = 1 . The concentration cio(TI) refers to solute i when that solute adsorbs singly from a solution at the same T and TI as that of the mixture .
In a dilute liquid mixture at constant temperature, µil is a function only of Ci , the concentration of solute i, in that mixture. Therefore, the right hand side of equation  becomes l# (T) +RT Ln C. µi l  Equations  to  lead to the second relationship of the IAS Model for calculating multi-solute equilbria Evaluation of Spreading Pressure (n)  In order to use Equation  to , n values must be known for the various singly-adsorbing solutes in the mixture.
The n values can be evaluated from equation (  1  The quantities of qio can be obtained from single solute isotherms, and can assume any of previously defined mathematical forms f(C~) 1  Total qT can be calculated from Eq. (2-32) The adsorption of each solute in a mixture can then be calculated from Eq. (2-28)  If the activated carbon dosage is represented by m  The equations  to  constitute a set of simultaneous equations from which the IAS Model calculations can be made.
There are a total of (5N+1) equat i ons.

Kinetics of Adsorption
Adsorption equilibrium is considered to be the most important factor for evaluating the feasibility of using activated carbon.
If enough time is allowed equilibrium will be achieved. However, for engineering applications, the rate of adsorption is a primary concern. Several distinct resistances to mass transfer may limit the adsorption rate and without detailed analysis coupled with subsequent experimental measurements, it is not always obvious which resistance is the rate limiting step . Associated with activated carbon adsorption, whether in a finite batch system or in a packed bed, there are two major rate limiting steps which include film transfer resistance and effective intraparticle resistance. Intraparticle resistance is divided into surface diffusion and pore diffusion.

Film Transfer Resistance •
The simplest theory that describes interfacial mass transfer is called the film theory. In this theory a hydrodynamic film surrounds the activated carbon pellet so that transfer of adsorbate occurs through the film. A mass balance with the assumption of a linear driving force across the film yields the following equation:  in which q = Adsorbed phas e concentration over the particle. They corrected the earlier correlations for axial dispersion and developed a correlation applicable to Reynold numbers ranging between 3-10,000. Their reevaluation of gas-phase data was considerably higher than those obtained under the assumption of zero fluid effective dispersion coefficient.
Levenspie158 presented an analytical solution that includes dispersion for calculating the film transfer coefficient from the immediate breakthrough concentration in a minicolumn study . In a similar approach Weber and Liu59 developed a method for evaluating both kf and Ds simultaneously in minicolumn such that the column i s sufficiently short not to contain the wavefront of sorbent-sorbate system.

Intraparticle Transfer Resistance
Intraparticle resistance is a lum p-sum term reflecting the  0 (for all t)  with this model, the pore diffusion only occurs within intraparticle pore voids and adsorbing on vacant pore wall sites.
Adsorbed molecules can resume pore diffusion only by first desorbing.
For the case of surface diffusion where surface flux predominates and pore flux can be neglected a material balance on a spherical particle with constant effective surface diffusion coefficient will yield: (2-53) The corresponding initial and boundary condition are q o at t 0 (0 <r <R)  aql dr r=O 0 (2-55)  where Ds(q) Surface diffusion coefficient as a function of the solid phase uptake Cs = Adsorbate concentration at the interface of the carbon particle M/L3 The previous mathematical relationships d is cussed the kinetics of adsorption of a single spherical particle exposed to a solute concentration at the particle's external surface. These mathemat i cal relationships can be used to describe adsorption in completely mixed batch reactors. For more practical engineering applications, the prediction of breakthrough curves from the basic kinetics of equilibrium data is a prerequisite, since it provides the right approach for evaluating the dynamic capacity of packed beds.
The differential liquid phase mass balance equation for an adsorption column and for spherical adsorbent particle within such a column can provide the mathematical distribution to describe the dynamic behavior of a packed bed. If the flow pattern can be reprresented as an axially dispersed plug flow the main balance will yield: The equilibrium relationship is expressed by the Freundlich Equations (2-75)

Development of Predictive Models
Predictive mathematical models can save substantial efforts in the initial planning stages of full-scale GAC treatment design.
The successful models should be able, once provided with the necessary input data, to predict the breakthrough curve characteristics of specific compound(s) of interest, as well as the impact of process variables on the process performance so that the optimum scheme can be identified . Batch reactor models and packed bed models that have been applied in this study were developed at Michigan Technological University81 and modified over a period of 12 years. The models were tested on multicomponent mixtures in clean background water. No attempt has been made to test these models for adsorption prediction in complex background mixtures or to verify them on large scale pilot plants treating complex wastewater prior to this study.

. 1 Batch Reactor Predictive Models
The batch reactor predictive models were developed, based on two intraparticle diffusion resistances. The first resistance, homogeneous surface diffusion resistance, is modeled by BHSDM.
The PSDM model considers that the pore and surface diffusion resistances act in parallel. Both numerical models include the following assumptions: 1 . External fluid film resistances exist .

RETURN TO MAIN PROGRAM AND PRINT OUT TIME T AND C
Similar to batch models, the packed bed predictive models incorporate either homogenous surface diffusion mechanism or homongenous and pore diffusion mechanism acting in parallel.
They also involves two flow schemes . The first flow scheme is plug flow where the axial dispersion term -De a2c/ az2 is neglected and the second flow is a scheme dispersed flow scheme in which the dispersion term is included in the axial transport of adsorbate in the bed.
The flow diagram81 for the packed bed model is shown in

Properties of Activated Carbon
The granular activated carbon (GAC) used in this study is Calgon's Filtrasorb 400 (Calgon Corporation, Pittsburg, PA); a bituminous based carbon activated with an oxygen-nitrogen mixture. Table 3-1 presents Filtrsorb 400's major properties as adopted from the manufacturer's data. One of the important properties of GAC that is usually not considered in presenting the activated carbon specifications is the pore radii and their distribution.
Since the molecular weight of the adsorbate affects the affinity of adsorption through surface attraction, higher molecular weight compounds possess a higher surface attraction. However, this may ~7 1.

Properties of Specific Organics
All chemicals were ultra pure grades and were obtained from Aldrich Chemical Company. The GC analysis of the base-neutral and acid extractables was performed using a Tracor Model 565 LSC-2 gas chromatograph equipped with split-splitless injection pore and connected to a Spectra-Physics SP-4270 integrator. The identification of the specific organics tested as well as their concentration was based on their retention times and peak area which have been determined experimentally by the analysis of standard reference materials.
The gas chromatograph operating conditions are presented in Table   3-3. Quality control ampule samples as received from the EPA A photograph of the contactor is shown in Figure 3   The pH of the solution was initially adjusted to 7 . The pH was measured after each sampling and only varied within the range of ~ 0 . 2 pH unit. A blank solution was initially prepared without the addition of carbon, and allowed to agitate until an equilibrium concentration was reached . The compounds 2 , 4-dimethylphenol and fluorene did not vary with time, however, naphthalene reached an equilibrium initial concentration after 26 hours. The Freundlich isotherm was used to calculate the required 50 x 60 mesh carbon dosage to achieve a 50% reduction in the initial solute concentration. This calculated carbon mass was then introduced into the reactor at t=O. Samples were then periodically withdrawn until equilibrium was reached, and then the samples were analyzed for solute concentration as described before. The total volume of all samples withdrawn was less than 15% of the reactor capacity.

Minicolumn Study Procedure
The minicolumn studies were conducted in 1 cm ID glass columns. In order to support the carbon in the column, a 1 cm  identified. An adsorption isotherm study is the most practical approach to characterizing the adsorption relationship. By performing a single-solute adsorption isotherm in ultrapure water background the competitive background effects can be eliminated.
By next performing the adsorpt i on isotherms in a mixture of complex background wastewater the competition effect due to background can be isolated.
For this study, adsorption isotherms were conducted at a pH of 6.8 and a temperature of 22°c which are typical seconda r ywastewater effluent characteristics in summer, the time planned for running the verification step using the pilot plant.
The pH can affect the results of isotherm study in various ways: 1) There is a close relationship between adsorbability and solubility. The less soluble a material is, the more likely that this material will be adsorbed.
2) The pH of the liquid will affect the dispersibility of the pulvarized carbon suspension. At pH levels below 7, a carbon suspension tends to agglomerate into larger floes which will enhanc e settling and filtration of the carbon particles.
The time required to achieve isotherm equilibrium has been reported differently in various research studies. An equilibrium time as low as 2 hours, at 22°c was reported by Dobbs and Cohen90.
By increasing the equilibrium time to about 2 days only increased the equilibrium capacity by 10 percent . The temperature at which the isotherm is conducted also appears to affect the equilibrium time52, with the decrease of temperature, the equilibrium time required to achieve the same percentage of equilibrium will increase. This is attributed to the decrease in the rate of     Naphthalene (0.481 g/g-C), pyrene (0.336 g/g-C) and bis(2-ethylhexyl) phthalate (0.057 g/g-C). This order is not comparable with the order resulting from the Freundlich model. Another isotherm analysis approach adopted by Manes93,94 and based also on Polanyi's theory was considered for isotherm analysis in this study. The difference between Dubinin-Polany's approach and Polany's approach as adopted by Manes is that Manes  Statistical values were obtained through the following formulations.

Relative Error (RE)
Mathematically defined as RE Ix -c I (4)(5) x This simple comparison between observed and predicted values is very rough and the statistics in it cannot recognize the variab i ity of data. However, this relationship can provide some measure of the model adequacy when other statistical parameters such as the media and standard deviation of the relat i ve sample are evaluated.

Root Mean Square Error (RMSE)
Mathematically defined as (4)(5)(6) This type of analysis provides a direct measure of model error.
The main disadvantage of this test is that it does not readily lend itself to pooling across variables to assess overall model credibility.

Analys i s of Variance in Linear Regression.
Accor ding to Bethea et a1.96, the proposed functional relationship between the observed and predicted data is where y and X are the observed and predicted values , respectively E is a random error, and the parameters $ 0 and $1 are the regression coefficients. The means square due to error, MSE=SSE/(n-2), is an unbiased estimator for o2 because the expected MSE = o2 regardless of whether or not the hypothesis Ho : $ 1 = o is true. It can be shown that the expected value of the mean square due to regression MSR = SSR/1, is a biased estimator for 0 2 unless $ 1 = 0 . This can be shown by These two expected mean squares suggest the use of the ratio It can be shown that T2 is an F-statistics with and n-2 degrees of freedom so that using F is equivalent to using T or T2.
Theoretically a value for $ 1 = 1 can be considered which is the slope of the regression line. In order to test whether or not the proposed s 1 is valid, the T statistics in the equation: sl (4)(5)(6)(7)(8)(9)(10) By checking that the value of T is less than the tabular of T o 95 the hypothesis that S1 1 ,n, . • 1 can be accepted or rejected within 95% probability.
The following method which was developed by Leggel and Williams97 was used in comparing the various models tested in this study. Their method is specific for evaluating a model's reliability or comparing between models. A reliability index k was defined which can be determined from set x1 ,x2, . .. ,xn of model prediction and a corresponding set y 1 ,y2, ... ,yn of observations. Applying their method one can express model predictibility as an "accurate within a factor of k," that is, the closer k is to one the better the model predictability is

Competitive Adsorption in Ultrapure Water Background
The competitive interaction of the preselected priority organics was examined in three combination sets based on their solubility in water (high, low and higha nd low solubilities combined). This approach is reasonable since for many compounds the solubility of the compound may indicate its adsorbability, that is the lower the solubility of a compound the higher its adsorbability. This general observation is confirmed by the results of the single-solute study discussed previously.
However, there are other factors such as a compound's molecular weight, polarity, refractive index and others which can influence a compound's adsorption on carbon.
In order to test the competition prediction by the IAS model using the modified Freundlich isotherm parameters, a bottle point isotherm study was conducted for each of the prementioned sets.
The raw data are presented in Appendix C in Table C-1 to Table   c

Competitive Adsor ption in Complex Wastewater Background
In this part of the study the total background effect in a complex mixture composition was treated as a single compound and was represented by the dissolved organic carbon (DOC) content.
This approach allowed the evaluation of the competitive interaction between any single solute or specific group of solutes of interest in the presence of a complex background mixture. The rational of this approach: 1 · The evaluation of adsorption capacity and the diffusion rate of specific compounds in the presence of a complex background is of great importance for actua l process design for wastewater treatment.

Due to cost const r aints and the present limitations in
analytical methodology it is almost impossible to completely identify the composition of a wastewater.
3. The use of a fictive or surrogate background component like DOC can permit the use of available isotherm predictive models without the necessity of a major alternation.

The selection of a DOC parameter to represent the lump-sum
background effect is significant for the following reasons: 1) DOC is easy to measure and insensitive to interferences commonly present in a complex wastewater, 2) only organic compounds have high affinity for activated carbon, therefore the presence of any other inorganic compounds such as heavy metals which are generally less adsorbable will not have a significant competitive background effect on adsorption in comparison to the DOC component.
The general procedure followed for conducting the isotherm tests was discussed previously in Section 3. However, a more detailed explanation of the procedures will be discussed in the           7.4 (4)(5)(6)(7)(8)(9)(10)(11) where DL Molecular diffusion of the compound, cm2/sec The por e diffus i on coefficient (Dp) was set equal to the mole c ular diffusion coefficient (DL) times the void fraction of the activat ed carbon Czp) .
The previously discussed coefficients we r e calculated for each of the compounds studied . Their values are presented in
The four prediction models were tested for predicting the    83.3 experimental data. Raw data are presented in Appendix E Table   E-1 Table E To substantiate its validity, its predictive capability was evaluated by statistical linear regression and its sensitivity was evaluated for the major mass transfer coefficients and isotherm constants. The results of the statistical linear regression tests are presented in Table 4-11.
The R-square, STD ERR and Root MSE values indicate a very good ability of PPSDM to predict the experimental data . However, the t-test theory indicated in all cases of intercept and slope a rejection outcome within 90% probability. The experimental data does not follow a pure normal distribution. By using C/C 0 which is less than one in all cases, very small standard error values  ....  a r e produced which in turn can give high t-test values that more likely will fail this test. The reliability index procedure as discussed in Section 3 can give a much better basis of comparison.
The procedure discussed in section 3 was originally designed for correlating various models for the same type of data sets and its application in this kind of study is more appropriate. The closer the Kg and Ks va l ues are to one the better correlation will be. The results of the reliability index parameters are presented i n Table 4    . .

Background Effects on Minicolumn Prediction
A minicolumn study was conducted to evaluate the impact of  can be visually observed that pore diffusion models gave a better  The flow to the packed beds was set to provide an average EBCT of  The pilot plant backwash system was arranged in a way to provide clean water for the dual sand filter and to use the actual wastewater passing through the sand filter to backwash the GAC columns. This sequence provided minimum disturbance of the GAC beds and reduced the backwash frequency of the sand filter to once every 24 hours and the GAC column to once every two to three days.
In the case of 3.56 minutes EBCT, a much better effluent profile prediction was produced. Statistical analysis of observed versus predicted results are presented in Table 4 (2-ethylhexyl) phthalate. Apart from the compound bis (2-ethylhexyl) phthalate, an increase in adsorption capacity was observed compared to the decrease in solubility of the compound in water. The deviation of bis (2-ethylhexyl) phthalate was attributed to its relative high molecular volume, however, there was no correlation between adsorption capacities and molecular volumes among the rest of the compounds.

3.
Dubinin-Polanyi adsorption isotherm worked well for fitting isotherm data. However it was unsuccessful as a predictive model when tested in · a bi-solute system. This is attributed to: lack of good solubility data and lack of good physical property data which could characterize adsorption behavior . I. Cleaning .procedure All glasswari used in the isotherm studies was cleaned from any trace organics according to the following procedure: a.
All glassware was washed with a free phosphate laboratory cleaning agent .
b . The washed classware was then rinsed with ultra pure water which was prepared in the following manner: 1 . Distilled water produced using Wheaton autostill -5 distillation unit.
2. The distilled water was passed through activated carbon column to remove traces of organics .
3. The organic free water was then filtered using millipore membrane filter (size 0 . 45 µ) to remove microorganisms .

II. Preparation of powdered activated carbon
As the time to equilibrium is dependent on activated carbon