Tài liệu ADSORPTION OF ORGANIC COMPOUNDS doc

Chap-The removal of organic compounds by adsorption on activated carbon is veryimportant in water purification and therefore is the primary focus of this chapter... musty odor compound,

CHAPTER 13 ADSORPTION OF ORGANIC

COMPOUNDS

Vernon L Snoeyink, Ph.D.

Ivan Racheff Professor of Environmental Engineering

Department of Civil and Environmental Engineering

University of Illinois at Urbana-Champaign

Adsorption of a substance involves its accumulation at the interface between twophases, such as a liquid and a solid or a gas and a solid The molecule that accumu-

lates, or adsorbs, at the interface is called an adsorbate, and the solid on which adsorption occurs is the adsorbent Adsorbents of interest in water treatment

include activated carbon; ion exchange resins; adsorbent resins; metal oxides,hydroxides, and carbonates; activated alumina; clays; and other solids that are sus-pended in or in contact with water

Adsorption plays an important role in the improvement of water quality vated carbon, for example, can be used to adsorb specific organic molecules thatcause taste and odor, mutagenicity, and toxicity, as well as natural organic matter(NOM) that causes color and that can react with chlorine to form disinfection by-products (DBPs) NOM is a complex mixture of compounds such as fulvic andhumic acids, hydrophilic acids, and carbohydrates The aluminum hydroxide and fer-ric hydroxide solids that form during coagulation will also adsorb NOM Adsorption

Acti-of NOM on anion exchange resins may reduce their capacity for anions (see ter 9), but ion exchange resins and adsorbent resins are available that can be used forefficient removal of selected organic compounds Calcium carbonate and magne-sium hydroxide solids formed in the lime softening process have some adsorptioncapacity, and pesticides adsorbed on clay particles can be removed by coagulationand filtration (Chapters 6 and 8)

Chap-The removal of organic compounds by adsorption on activated carbon is veryimportant in water purification and therefore is the primary focus of this chapter A

13.1

study conducted by two committees of the AWWA showed that approximately 25percent of 645 United States utilities, including the 500 largest, used powdered acti-vated carbon (PAC) in 1977 (American Water Works Association, 1977) In 1986, 29percent of the 600 largest utilities reported using PAC (American Water WorksAssociation, 1986), predominantly for odor control More attention is being givennow to granular activated carbon (GAC) as an alternative to PAC GAC is used incolumns or beds that permit higher adsorptive capacities to be achieved and easierprocess control than is possible with PAC The higher cost for GAC can often be off-set by better efficiency, especially when organic matter must be removed on a con-tinuous basis GAC should be seriously considered for water supplies when odorouscompounds or synthetic organic chemicals of health concern are frequently present,when a barrier is needed to prevent organic compounds from spills from enteringfinished water, or in some situations that require DBP control GAC has excellentadsorption capacity for many undesirable substances and it can be removed fromthe columns for reactivation when necessary The number of drinking water plantsusing GAC, principally for odor control, increased from 65 in 1977 (American WaterWorks Association, 1977) to 135 in 1986 (Fisher, 1986); in 1996, there were approxi-mately 300 plants treating surface water and several hundred more treating contam-inated groundwater The promulgated as well as proposed DBP regulations willdrive many utilities to consider GAC for removal of organic compounds in the next

10 years GAC is also used as a support medium for bacteria in processes to cally stabilize drinking water before distribution

biologi-This chapter also covers the use of ion exchange and adsorbent resins for theremoval of organic compounds Removal of inorganic ions by ion exchange resinsand activated alumina is discussed in Chapter 9

ADSORPTION THEORY

Adsorption Equilibrium

Adsorption of molecules can be represented as a chemical reaction:

A +B ⇔A⋅B where A is the adsorbate, B is the adsorbent, and A⋅B is the adsorbed compound.

Adsorbates are held on the surface by various types of chemical forces such ashydrogen bonds, dipole-dipole interactions, and van der Waals forces If the reac-tion is reversible, as it is for many compounds adsorbed to activated carbon,molecules continue to accumulate on the surface until the rate of the forward reac-tion (adsorption) equals the rate of the reverse reaction (desorption) When thiscondition exists, equilibrium has been reached and no further accumulation willoccur

Isotherm Equations. One of the most important characteristics of an adsorbent isthe quantity of adsorbate it can accumulate The constant-temperature equilibrium

relationship between the quantity of adsorbate per unit of adsorbent qeand its

equi-librium solution concentration Ce is called the adsorption isotherm Several

equa-tions or models are available that describe this function (Sontheimer, Crittenden,and Summers, 1988), but only the more common equations for single-solute adsorp-tion, the Freundlich and the Langmuir equations, are presented here

The Freundlich equation is an empirical equation that is very useful because itaccurately describes much adsorption data This equation has the form

and can be linearized as follows:

The parameters qe(with units of mass adsorbate/mass adsorbent, or mole

adsor-bate/mass adsorbent) and Ce(with units of mass/volume, or moles/volume) are the

equilibrium surface and solution concentrations, respectively The terms K and 1/n are constants for a given system; 1/n is unitless, and the units of K are determined by the units of qe and Ce Although the Freundlich equation was developed to empiri-

cally fit adsorption data, a theory of adsorption that leads to the Freundlich equationwas later developed by Halsey and Taylor (1947)

The parameter K in the Freundlich equation is related primarily to the capacity

of the adsorbent for the adsorbate, and 1/n is a function of the strength of tion For fixed values of Ce and 1/n, the larger the value of K, the larger the capacity

adsorp-q e For fixed values of K and Ce, the smaller the value of 1/n, the stronger is the adsorption bond As 1/n becomes very small, the capacity tends to be independent of

C e and the isotherm plot approaches the horizontal level; the value of qethen is

essentially constant, and the isotherm is termed irreversible If the value of 1/n is large, the adsorption bond is weak, and the value of qechanges markedly with small

changes in Ce.

The Freundlich equation cannot apply to all values of Ce , however As C e increases, for example, qeincreases (in accordance with Equation 13.1) only until the

adsorbent approaches saturation At saturation, qeis a constant, independent of

fur-ther increases in Ce, and the Freundlich equation no longer applies Also, no

assur-ance exists that adsorption data will conform to the Freundlich equation over allconcentrations less than saturation, so care must be exercised in extending the equa-tion to concentration ranges that have not been tested

The Langmuir equation,

where b and qmaxare constants and qe and Ceare as defined earlier, has a firm

theo-retical basis (Langmuir, 1918) The constant qmaxcorresponds to the surface

concen-tration at monolayer coverage and represents the maximum value of qethat can be

achieved as Ce is increased The constant b is related to the energy of adsorption and

increases as the strength of the adsorption bond increases The Langmuir equationoften does not describe adsorption data as accurately as the Freundlich equation

The experimentally determined values of qmaxand b often are not constant over the

concentration range of interest, possibly because of the heterogeneous nature of theadsorbent surface (a homogeneous surface was assumed in the model develop-ment), lateral interactions between adsorbed molecules (all interaction wasneglected in the model development), and other factors

Factors Affecting Adsorption Equilibria. Important adsorbent characteristicsthat affect isotherms include surface area, pore size distribution, and surface chem-

FIGURE 13.1 Pore size distributions for different

acti-vated carbons (Source: Lee, Snoeyink, and Crittenden,

1981.)

istry The maximum amount of adsorption is proportional to the amount of surfacearea within pores that is accessible to the adsorbate Surface areas range from a fewhundred to more than 1500 m2/g, but not all of the area is accessible to aqueousadsorbates The range of pore size distributions in an arbitrary selection of GACs isshown in Figure 13.1 A relatively large volume of micropores (pores less than 2 nm

diameter d) (Sontheimer, Crittenden, and Summers, 1988) generally corresponds to

a large surface area and a large adsorption capacity for small molecules, whereas alarge volume of mesopores (2 <d<50 nm) and macropores (d>50 nm) is usuallydirectly correlated to capacity for large molecules The fulvic acid isotherms in Fig-ure 13.2 are for the same activated carbons whose pore size distributions are shown

in Figure 13.1 Note that the activated carbons that have a relatively small volume

of macropores also have a relatively low capacity for the large fulvic acid molecule.Lee et al (1981) showed that the quantity of humic substances of a given size thatwas adsorbed was correlated with pore volume within pores of a given size The rel-ative positions of the isotherms for the activated carbons in Figure 13.1 might beentirely different than those in Figure 13.2 if the adsorbate were a small molecule,such as a phenol, which can enter pores much smaller than those accessible to ful-vic acid Summers and Roberts (1988b) showed that if the amount adsorbed wasnormalized for the available surface area, the differences in adsorption capacity ofdifferent carbons for a humic acid could be attributed to the surface chemistry ofthe carbon

The surface chemistry of activated carbon and adsorbate properties also canaffect adsorption (Coughlin and Ezra, 1968; Gasser and Kipling, 1959; Kipling andShooter, 1966; Snoeyink and Weber, 1972; Snoeyink et al., 1974) Several researchers(Coughlin and Ezra, 1968; Gasser and Kipling, 1959; Kipling and Shooter, 1966)demonstrated that extensive oxidation of carbon surfaces led to large decreases inthe amounts of phenol, nitrobenzene, benzene, and benzenesulfonate that could beadsorbed Oxidation of the activated carbon surface with aqueous chlorine was alsofound to increase the number of oxygen surface functional groups and correspond-ingly to decrease the adsorption capacity for phenol (Snoeyink et al., 1974) Thus,oxygenating a carbon surface decreases its affinity for simple aromatic compounds

The tendency of a molecule to adsorb is a function of its affinity for water as pared to its affinity for the adsorbent Adsorption onto GAC from water, for exam-ple, generally increases as the adsorbate’s solubility decreases (Weber, 1972) As amolecule becomes larger through the addition of hydrophobic groups such as

com-CH2, its solubility decreases and its extent of adsorption increases as long asthe molecule can gain entrance to the pores When an increase in size causes themolecule to be excluded from some pores, however, adsorption capacity maydecrease as solubility decreases As molecular size increases, the rate of diffusionwithin the activated carbon particle decreases, especially as molecular sizeapproaches the particle’s pore diameter

The affinity of weak organic acids or bases for activated carbon is an importantfunction of pH When pH is in a range at which the molecule is in the neutral form,adsorption capacity is relatively high When pH is in a range at which the species isionized, however, the affinity for water increases and activated carbon capacityaccordingly decreases Phenol that has been adsorbed on activated carbon at pHbelow 8, where phenol is neutral, can be desorbed if the pH is increased to 10 orabove, where the molecule is anionic (Fox, Keller, and Pinamont, 1973) If adsorp-tion occurs on resins by means of the ion exchange mechanism, the specific affinity

of the ionic adsorbate for charged functional groups may also cause good removal.The inorganic composition of water also can have an important effect on theextent of NOM adsorption, as shown in Figure 13.3 for fulvic acids (Randtke andJepsen, 1982) After 70 days, a small GAC column was nearly saturated with fulvicacid Addition of CaCl2at this point resulted in a large increase in adsorbability offulvic acid, as reflected in the reduced column effluent concentration After 140days, elimination of the CaCl2resulted in desorption of much of the fulvic acid Cal-cium ion apparently associates (complexes) with the fulvic acid anion to make ful-vic acid more adsorbable (Randtke and Jepsen, 1982; Weber, Voice, and Jodellah,1983) Presumably many other divalent ions can act in similar fashion, but calcium

is of special interest because of its relatively high concentration in many naturalwaters Similar effects are expected for other anionic adsorbates, but salts are notexpected to have much effect on the adsorption of neutral adsorbates (Snoeyink,Weber, and Mark, 1969)

FIGURE 13.2 Adsorption isotherms for peat fulvic

acid (Source: Lee, Snoeyink, and Crittenden, 1981.)

Inorganic substances such as iron, manganese, and calcium salts or precipitatesmay interfere with adsorption if they deposit on the adsorbent Pretreatment toremove these substances, or to eliminate the supersaturation, may be necessary ifthey are present in large amounts.

Adsorption isotherms may be determined for heterogeneous mixtures of pounds using group parameters such as total organic carbon (TOC), dissolvedorganic carbon (DOC), chemical oxygen demand (COD), dissolved organic halogen(DOX), UV absorbance, and fluorescence as a measure of the total concentration ofsubstances present Because the compounds within a mixture can vary widely intheir affinity for an adsorbent, the shape of the isotherm will depend on the relativeamounts of compounds in the mixture For example, isotherms with the shape shown

com-in Figure 13.4 are expected if some of the compounds are nonadsorbable and someare more strongly adsorbable than the rest (Randtke and Snoeyink, 1983) Thestrongly adsorbable compounds can be removed with small doses of adsorbent and

yield large values of qe In contrast, the weakly adsorbable compounds can only be removed with large doses of adsorbent that yield relatively low values of qe The nonadsorbable compounds produce a vertical isotherm at low Cevalues In contrast

to single-solute isotherms, the isotherm for a heterogeneous mixture of compoundswill be a function of initial concentration and the fraction of the mixture that isadsorbed The relative adsorbabilities of compounds within a mixture have animportant effect on the performance of adsorption columns The nonadsorbablefraction cannot be removed regardless of the column design, whereas the stronglyadsorbable fraction may cause the effluent concentration to slowly approach theinfluent concentration

Competitive Adsorption in Bisolute Systems. Competitive adsorption is tant in drinking water treatment because most compounds to be adsorbed exist insolution with other adsorbable compounds The quantity of activated carbon orother adsorbent required to remove a certain amount of a compound of interest

impor-FIGURE 13.3 Effects of calcium chloride addition and withdrawal on column formance (pH = 8.3; TOC = 5.37 mg/L, peat fulvic acid buffer =1.0 mM NaHCO3 ).

per-(Source: Randtke and Jepsen, 1982.)

from a mixture of adsorbable compounds

is greater than if adsorption occurs out competition, because part of theadsorbent’s surface is utilized by thecompeting substances

with-The extent of competition on activatedcarbon depends upon the strength ofadsorption of the competing molecules,the concentrations of these molecules,and the type of activated carbon Someexamples illustrate the possible mag-nitude of the competitive effect Jain and Snoeyink (1973) showed that as

p-bromophenol (PBP) equilibrium

con-centration increased from 10−4to 10−3M

FIGURE 13.4 Nonlinear isotherm for a

het-erogeneous mixture of organic compounds.

(Source: Randtke and Snoeyink, 1983.)

FIGURE 13.5 Breakthrough curves for sequential feed of DMP and DCP to

a GAC adsorber (C0 =0.990 mmol/L, C02 = 1.02 mmol/L, EBCT = 25.4 s).

(Source: W E Thacker, J C Crittenden, and V L Snoeyink, 1984 “Modeling of

Adsorber Performance: Variable Influent Concentration and Comparison of

Adsorbents,” Journal Water Pollution Control Federation 56: 243 Copyright ©

Water Environment Federation, reprinted with permission.)

saturated with DMP (Thacker, Crittenden, and Snoeyink, 1984) Similar occurrenceshave been observed in full-scale GAC systems Effluent concentrations in excess ofinfluent concentrations can be prevented through careful operation Crittenden et

al (1980) showed that the magnitude of the displacement decreased when the value

of Ceff/Cinfwas lowered at the time the second compound was introduced Thus, areasonable strategy to prevent the occurrence of an undesirable compound at a con-centration greater than the influent is (1) to monitor the column for that compoundand (2) to replace the activated carbon before complete saturation at the influent

concentration occurs (i.e., before Ceff=Cinf)

A number of isotherm models have been used to describe competitive tion A common model for describing adsorption equilibrium in multiadsorbate sys-tems is the Langmuir model for competitive adsorption, which was first developed

adsorp-by Butler and Ockrent (1930) and which is presented in the fourth edition of thisbook (Snoeyink, 1990) This model is based on the same assumptions as the Lang-muir model for single adsorbates Jain and Snoeyink (1973) modified this model toaccount for a fraction of the adsorption taking place without competition This canhappen if the adsorbates have different sizes and only the smaller adsorbate canenter the smaller pores (Pelekani and Snoeyink, 1999), or if some of the surfacefunctional groups adsorb one compound but not the other Other models that can beused to describe and predict competitive effects are the Freundlich-type isotherm ofSheindorf, Rebhun, and Sheintuck (1981) and the ideal adsorbed solution theory ofRadke and Prausnitz (1972) described in the next section The latter has proven to

be applicable to a large number of situations

Competitive Adsorption in Natural Waters. Adsorption of organic compounds attrace concentrations from natural waters is an important problem in water purifica-tion Essentially all synthetic organic chemicals that must be removed in water treat-ment by adsorption must compete with natural or background organic matter foradsorption sites The heterogeneous mixture of compounds in natural watersadsorbs on activated carbon and reduces the number of sites available for the tracecompounds, either by direct competition for adsorption sites or by pore blockage(Pelekani and Snoeyink, 1999) The amount of competition and the capacity for thetrace compound depend on the nature of the background organic matter and its con-centration, as well as the characteristics of the activated carbon Also important isthe concentration of the trace compound, because this concentration affects howmuch of this compound can adsorb on the carbon For example, Figure 13.6 showsthat the adsorption capacity of 2-methylisoborneol (MIB), an important earthy/

FIGURE 13.6 Effect of initial concentration on MIB capacity in Lake

Michigan water (Source: Gillogly et al., 1998b.)

musty odor compound, is lower in natural water than in distilled water, and that thiscapacity is further reduced as initial concentration decreases (Gillogly et al., 1998b).

It is important to have a procedure to predict capacity as a function of initial centration, because the capacity of activated carbon depends in such an importantway on initial concentration and because the concentrations of trace organic com-pounds vary widely in natural waters The ideal adsorbed solution theory (IAST)can be used for this purpose The following two equations, based on the IAST asdeveloped by Radke and Prausnitz (1972) and modified by Crittenden et al (1985)

con-to include the Freundlich equilibrium expression, describe equilibrium in a solute system,

two-C1,0−q1C c−  n1

C2,0−q2C c−  n2

where q1and q2=equilibrium solid phase concentrations of compounds 1 and 2

q1=(C1,0−C 1,e )/Cc and q2=(C2,0−C 2,e )/Cc

C1,0and C2,0=initial liquid phase concentrations of compounds 1 and 2

C 1,e and C 2,e=equilibrium concentrations of compounds 1 and 2

K1and K2=single-solute Freundlich parameters for compounds 1 and 2

1/n1and 1/n2=single-solute Freundlich exponents for compounds 1 and 2

C c=carbon doseThese equations show the relationship between the initial concentration of eachadsorbate, the amount of adsorbed compound per unit weight of carbon, and thecarbon dose The Freundlich parameters are derived from single-solute tests inorganic-free water

In natural waters, the organic matter present is a complex mixture of many ferent compounds; representing each of these compounds, even if they could beidentified, would be computationally prohibitive Several researchers have modeledNOM adsorption by defining several fictive components that represent groups of

dif-compounds with similar adsorption characteristics, as expressed by Freundlich K and n values (Sontheimer, Crittenden, and Summers, 1988) Extending Equations 13.4 and 13.5 to N components yields

where N=number of components in the solution

C i,0=initial liquid-phase concentration of compound i

C c=carbon dose

q i=equilibrium solid-phase concentration of compound i

n i and Ki=single-solute Freundlich parameters for compound i

These equations can be solved simultaneously to determine the concentrations foreach component assumed to be in solution

Crittenden et al (1985) used this fictive component approach to describe theadsorption of a target compound in the presence of NOM With a single-soluteisotherm of the target compound and experimental results from isotherms measured

FIGURE 13.7 EBC model results for atrazine isotherms in Illinois

ground-water (Source: Reprinted with permission from D R U Knappe et al 1998.

“Predicting the capacity of powdered activated carbon for trace organic

com-pounds in natural waters.” Environmental Science & Technology, 32: 1694–1698.

Copyright 1998 American Chemical Society.)

using the natural water, parameters for each of the fictive components were foundthrough a best-fit search procedure These results were then applied to describe theadsorption of other compounds in that water

The IAST was applied to the problem of trace organic adsorption in naturalwaters by Najm, Snoeyink, and Richard (1991) using a procedure that was subse-quently modified by Qi et al (1994) and Knappe et al (1998) These researchersassumed that the background organic matter that competed with the trace com-pound could be represented as a single compound, called the equivalent backgroundcompound (EBC) This approach involved the determination of the single-soluteisotherm for the trace compound, and an isotherm in natural water for the tracecompound at two different initial concentrations A search routine was used to find

the Freundlich parameters K and 1/n and the initial concentration C0for the EBCthat gave the observed amount of competition For example, Figure 13.7 showsisotherms determined for atrazine in organic-free water and in Illinois groundwater

at initial concentrations of 176 and 36 µg/L (Knappe et al., 1998) These data wereused to determine the following EBC characteristics:

KEBC>1.0 ×106(µmole/g)(L/µmole)1/n , 1/nEBC=0.648, C0,EBC=0.870 µmole/L

The K value for the EBC was arbitrary above 1.0 ×106(µmole/g)(L/µmole)1/n.These EBC parameters are specific for the type of carbon, the type and concentra-tion of background organic matter, and the type of synthetic organic chemical(SOC) They can be used in Equations 13.4 and 13.5 together with the initial con-centration of the trace compound and its single-solute Freundlich parameters to cal-

culate the surface coverage of trace compound as a function of carbon dose Ce Given the surface coverage q, the initial concentration C0, and the carbon dose Cc,

the equilibrium concentration of the trace compound can be calculated from the

equation q = (C0 − C e)/Cc This approach was used to determine the predicted

isotherm for atrazine at an initial concentration of 8.3 µg/L, shown in Figure 13.7,which compares very well with the measured data It is interesting to note that at anequilibrium concentration of 1 µg/L, there is a 63 percent reduction in capacity foratrazine as the initial concentration is reduced from 176 to 8.3 µg/L

An important modification of the EBC model was developed by Knappe et al.(1998), who found that the amount of adsorption on a unit mass of activated carbonwas directly proportional to the initial concentration of that trace compound in an

adsorption test if (1) the Freundlich exponents for the trace compound 1/n1and the

EBC 1/n2both fall between 0.1 and 1, as is generally the case, and (2) the solid-phaseconcentration of the background organic matter is much in excess of that of thetrace compound, which also is often true When these two approximations hold, theIAST model simplifies to

pro-of this result is that only one isotherm need be determined for low concentrations

of trace compound in a natural water This isotherm can be plotted as percentremaining versus carbon dose, as shown in Figures 13.8 and 13.9 for the same datashown in Figures 13.6 and 13.7, respectively In Figure 13.9, the atrazine data for

C0≤36 µg/L plot on a single percent remaining versus carbon dose line, and the

C0=176 µg/L line is only slightly higher All the MIB data in Figure 13.6 plot onone line in Figure 13.8

FIGURE 13.8 Percent MIB remaining as a function of PAC dose.

(Source: Gillogly et al., 1998b.)

Graham et al (1999) independently derived a different form of Eq 13.7 andapplied it using the EBC approach to adsorption of MIB and geosmin from four nat-

ural waters The same EBC parameters [K=1.35 (µg/mg)(L/µg)1/n and 1/n=0.20]were found to be applicable to both MIB and geosmin and, at different initial con-centrations (15 to 51 µg/L), to all four natural waters For three of the four naturalwaters, the EBC initial concentration was about 0.45 percent of the TOC initial con-centration

These results allow the following general procedure to be used to determineadsorption capacity for a trace compound in natural water:

1 Determine one adsorption isotherm for the trace compound in natural water at a

sufficiently low initial concentration

2 Plot the data on a log-log percent remaining versus carbon dose plot.

3 Use this isotherm plot to determine the carbon dose required for any desired

per-cent removal for any initial conper-centration that satisfies the assumptions

An important question now is what concentration is sufficiently low that theassumptions made in developing the percent remaining versus carbon dose plot arevalid Research to date has shown that MIB (Gillogly et al., 1998b) and geosmin(Graham et al., 1999) concentrations less than 1000 ng/L, and atrazine concentra-tions less than about 50 µg/L32, will give a satisfactory plot However, it is necessary

to expand this database for other trace compounds

Desorption. Adsorption of many compounds is reversible, which means that theycan desorb Desorption may be caused by displacement by other compounds, as dis-cussed previously, or by a decrease in influent concentration Both phenomena mayoccur in some situations An analysis of desorption by Thacker, Snoeyink, and Crit-tenden (1983) showed that the quantity of adsorbate that can desorb in response to

a decrease in influent concentration increased as (1) the diffusion coefficient of the

FIGURE 13.9 Removal efficiency as a function of PAC dose for the

adsorption of atrazine from Illinois groundwater (GW) (Source:

Re-printed with permission from D R U Knappe et al., 1998 “Predicting the

capacity of powdered activated carbon for trace organic compounds in

natural waters.” Environmental Science & Technology, 32: 1694–1698.

Copyright 1998 American Chemical Society.)

adsorbate increased, (2) the amount of compound adsorbed increased, (3) the

strength of adsorption decreased (e.g., as the Langmuir b value decreased or the Freundlich l/n value increased), and (4) the activated carbon particle size decreased.

Volatile organic compounds are especially susceptible to displacement because theyare weakly adsorbed and diffuse rapidly Summers and Roberts (1988a, b) haveshown that NOM only partially desorbs and for the desorbing fraction, the desorp-tion diffusivity is lower than that during adsorption

Adsorption Kinetics

Transport Mechanisms. Removal of organic compounds by physical adsorption

on porous adsorbents involves a number of steps, each of which can affect the rate

of removal:

1 Bulk solution transport Adsorbates must be transported from bulk solution tothe boundary layer of water surrounding the adsorbent particle The transportoccurs through diffusion if the adsorbent is suspended in quiescent water such as

a sedimentation basin, or through turbulent mixing such as during turbulent flowthrough a packed bed of GAC, or when PAC is being mixed in a rapid mix unit orflocculator

2 External (film) resistance to transport Adsorbates must be transported bymolecular diffusion through the stationary layer of water (hydrodynamic bound-ary layer) that surrounds adsorbent particles when water is flowing past them.The distance of transport, and thus the time for this step, is determined by theflow rate past the particle The higher the flow rate, the shorter the distance

3 Internal (pore) transport After passing through the hydrodynamic boundarylayer, adsorbates must be transported through the adsorbent’s pores to availableadsorption sites Intraparticle transport may occur by molecular diffusionthrough the solution in the pores (pore diffusion), or by diffusion along theadsorbent surface (surface diffusion) after adsorption takes place

4 Adsorption After transport to an available site, an adsorption bond is formedbetween the adsorbate and adsorbent This step is very rapid for physical adsorp-tion (Adamson, 1982) and as a result one of the preceding diffusion steps willcontrol the rate at which molecules are removed from solution If adsorption isaccompanied by a chemical reaction that changes the nature of the molecule, thechemical reaction may be slower than the diffusion step and thereby control therate of compound removal

The transport steps occur in series, so the slowest step, called the rate-limiting step,

will control the rate of removal In turbulent flow reactors, a combination of film fusion and pore diffusion very often controls the rate of removal for some of thetypes of molecules to be removed from drinking water Initially, film diffusion maycontrol the rate of removal, and after some adsorbate accumulates within the pore,pore transport may control the rate of removal The mathematical models of theadsorption process, therefore, usually include both steps

dif-Both molecular size and adsorbent particle size have important effects on therate of adsorption Diffusion coefficients, in particular, decrease as molecular sizeincreases, and thus longer times are required to remove the large-molecular-weighthumic substances than are needed for the low-molecular-weight phenols, for exam-ple Adsorbent particle size is also important because it determines the timerequired for transport within the pore to available adsorption sites If the rate of

adsorbate uptake is controlled by intraparticle diffusion, and the effective diffusioncoefficient is constant, the time to reach equilibrium is directly proportional to thediameter of the particle squared Calculations by Randtke and Snoeyink (1983) foractivated carbon illustrate these points For the low-molecular-weight dimethylphe-

nol, nearly 8 days is estimated for near-equilibrium (Cfinal= 1.01 Ce) of

2.4-mm-diameter activated carbon, but only about 25 min is required for 44-µm-diameteractivated carbon For very large-molecular-weight (approximately 50,000) humicacid, the 2.4-mm-diameter particle is expected to take much longer than a year toequilibrate, but only 2 days is required for the 44-µm-diameter particle Calcula-tions for 10,000-molecular-weight fulvic acid showed only about 25 percent satura-tion of 2.4-mm-diameter particles after 40 days of contact Thus, the smaller theparticle, the faster equilibrium is achieved in both column and complete-mixadsorption systems

Some conclusions that can be drawn from these observations are that (1) lar carbon should be pulverized for isotherm measurement, especially when thecapacity of large molecular weight compounds is to be determined (pulverizing doesnot affect the total surface available for adsorption if all of the GAC is pulverized)(Randtke and Snoeyink, 1983); (2) the smallest activated carbon, consistent withother process constraints such as head loss and loss during reactivation, should bechosen for the best kinetics; and (3) depending on the compound, all the capacity oflarge activated carbon particles in a column may not be used because the time inter-val between activated carbon replacements is not sufficient for equilibrium to beachieved

granu-Mass Transfer Zone and Breakthrough Curves for Packed Bed Reactors. Theregion of an adsorption column in which adsorption is taking place, the mass trans-

fer zone (MTZ), is shown in Figure 13.10a The activated carbon behind the MTZ has been completely saturated with adsorbate at Ce=C0, and the amount adsorbed

per unit mass of GAC is (qe)0 The activated carbon in front of the MTZ has not beenexposed to adsorbate, so solution concentration and adsorbed concentration areboth zero Within the MTZ, the degree of saturation with adsorbate varies from 100

percent (q=[qe]0) to zero The length of the MTZ, LMTZ, depends upon the rate ofadsorption and the solution flow rate Anything that causes a higher rate of adsorp-tion, such as a smaller carbon particle size, higher temperature, a larger diffusioncoefficient of adsorbate, and/or greater strength of adsorption of adsorbate (i.e., a

larger Freundlich K value), will decrease the length of the MTZ In some stances, LMTZwill be reduced sufficiently that it can be assumed to be zero, yielding

circum-the ideal plug-flow behavior, as shown in Figure 13.10b If LMTZis negligible, sis of the adsorption process is greatly simplified

analy-The breakthrough concentration CBfor a column is defined as the maximumacceptable effluent concentration When the effluent concentration reaches this

value, the GAC must be replaced The critical depth of a column Lcriticalis the depth

that leads to the immediate appearance of an effluent concentration equal to CB when the column is started up For the situation in which CBis defined as the mini-mum detectable concentration, the critical depth of an activated carbon column isequal to the length of the MTZ The length of the MTZ is fixed for a given set of con-

ditions, but Lcriticalvaries with CB The critical depth, the flow rate Q, and the area of the column A, can be used to calculate the minimum empty bed contact time

(EBCT) (EBCT =Q/V, where V is the bulk volume of GAC in the contactor):

Lcritical



Q/A

When CBis greater than the minimum detectable concentration, the critical depth is

less than LMTZand its value can be determined as shown in a later section (see ure 13.14 and related discussion)

Fig-The breakthrough curve is a plot of the column effluent concentration as a tion of either the volume treated, the time of treatment, or the number of bed vol-umes (BV) treated (BV =V B/V, where VBis the volume treated) The number of bedvolumes is a particularly useful parameter because the data from columns of differ-ent sizes and with different flow rates are normalized A breakthrough curve for asingle adsorbable compound is shown in Figure 13.11 The shape of the curve isaffected by the same factors that affect the length of the MTZ, and in the same way.Anything that causes the rate of adsorption to increase will increase the sharpness ofthe curve, while increasing the flow rate will cause the curve to “spread out” over a

func-larger volume of water treated The breakthrough curve will be vertical if LMTZ=0,

as shown in Figure 13.10b As shown in Figure 13.11, the breakthrough capacity,

defined as the mass of adsorbate removed by the adsorber at breakthrough, and thedegree of column utilization, defined as the mass adsorbed at breakthrough/massadsorbed at complete saturation at the influent concentration, both increase as therate of adsorption increases

The breakthrough curve can be used to determine the activated carbon usagerate (CUR), the mass of activated carbon required per unit volume of water treated:

FIGURE 13.10 Adsorption column MTZ (a) Column with MTZ (b) Column without MTZ.

CUR  = (13.9)Breakthrough curves are strongly affected by the presence of nonadsorbable com-pounds, the biodegradation of compounds in a biologically active column, slowadsorption of a fraction of the molecules present, and the critical depth of the col-umn relative to the length of the column Immediate breakthrough of adsorbable

compounds occurs if the LMTZis greater than the activated carbon bed depth pare curves A and B in Figure 13.12) Nonadsorbable compounds immediately

(com-appear in the column effluent, even when the carbon depth is greater than the LMTZ(compare curves B and C in Figure 13.12) Removal of adsorbable, biodegradablecompounds by microbiological degradation in a column results in continualremoval, even after the carbon is saturated with adsorbable compounds (see curve

D in Figure 13.12) If a fraction of compounds adsorbs slowly, the upper part of thebreakthrough curve will be similar to that produced by biodegradation but will

slowly approach Ceff/C0=1 Breakthrough curves shown in later sections (see ures 13.15 and 13.22, for example) also illustrate some of these effects

Fig-GAC ADSORPTION SYSTEMS

Characteristics of GAC

Physical Properties. A wide variety of raw materials can be used to make vated carbon (Hassler, 1974), but wood, peat, lignite, subbituminous coal, and bitu-minous coal are the substances predominately used for drinking water treatmentcarbons in the United States Both the physical and chemical manufacturing pro-cesses involve carbonization (conversion of the raw material to a char) and activa-tion (oxidation to develop the internal pore structure) With physical activation,

acti-carbonization, or pyrolysis, is usually performed in the absence of air at

tempera-tures less than 700°C, while activation is carried out with oxidizing gases such as

mass of GAC in column



volume treated to breakthrough VB

mass

volume

FIGURE 13.11 Adsorption column breakthrough curve.

steam and CO2at temperatures of 800 to 900°C Chemical activation combines bonization and activation steps Patents describing carbonization and activation pro-cedures are given by Yehaskel (1978).

car-Various characteristics of activated carbon affect its performance.* The particleshape of crushed activated carbon is irregular, but extruded activated carbons have asmooth cylindrical shape Particle shape affects the filtration and backwash properties

of GAC beds Particle size is an important parameter because of its effect on rate ofadsorption, as discussed previously Particle size distribution refers to the relativeamounts of different-size particles that are part of a given sample, or lot, of carbon, andhas an important impact on the filtration properties of GAC in GAC columns that areused both as filters to remove particles and as adsorbers (i.e., filter-adsorbers) (Graese,Snoeyink, and Lee, 1987) Commonly available activated carbon sizes are 12 ×40 and

8 ×30 US Standard mesh, which range in apparent diameter from 1.68 to 0.42 mm andfrom 2.38 to 0.59 mm, respectively The uniformity coefficient (see Chapter 8) is oftenquite large, typically on the order of 1.9, to promote stratification during backwashing.Commercially available activated carbons usually have a small percentage of materialsmaller than the smallest sieve and larger than the largest sieve, which significantlyaffects the uniformity coefficient Extruded carbon particles all have the same diame-ter, but vary in length There is no method comparable to the sieve analysis procedure

to characterize the distribution of lengths, however

FIGURE 13.12 Effect of biodegradation and the presence of nonadsorbable compounds

on the shapes of breakthrough curves.

* Note: Descriptions of the analytical procedures for testing activated carbon are given in ASTM dards (American Society for Testing Materials, 1988) available from ASTM, 1916 Race St., Philadelphia, PA

Stan-19103, as well as in AWWA Standards B604-96 and B600-66, available from AWWA, 6666 West Quincy Ave.,

The apparent density* of activated carbon is the mass of nonstratified dry

acti-vated carbon per unit volume of actiacti-vated carbon, including the volume of voidsbetween grains Typical values for GAC are 350 to 500 kg/m3(25 to 31 lb/ft3) Distin-

guishing between the apparent density and the bed density, backwashed and drained

(i.e., stratified, free of water) is important, however The former is a characteristic ofactivated carbon as shipped.The latter is about 10 percent less than the apparent den-sity and is typical of activated carbon during normal operation unless it becomesdestratified during backwashing The latter is an important parameter because itdetermines how much activated carbon must be purchased to fill a filter of given size

The particle density wetted in water is the mass of solid activated carbon plus the

mass of water required to fill the internal pores per unit volume of particle Its valuefor GAC typically ranges from 1300 to 1500 kg/m3(90 to 105 lb/ft3) and it determinesthe extent of fluidization and expansion of a particle of given size during backwash.Particle hardness is important because it affects the amount of attrition duringbackwash, transport, and reactivation In general, the harder the activated carbon,the less the attrition for a given amount of friction or impact between particles Acti-vated carbon hardness is generally characterized by an experimentally determinedhardness or abrasion number, using a test such as the ASTM Ball Pan Hardness Testthat measures the resistance to particle degradation upon agitation of a mixture ofactivated carbon and steel balls (American Society for Testing Materials, 1988) Therelationship between the amount of attrition that can be expected when activatedcarbon is handled in a certain way and the hardness number has not been deter-mined, however

Adsorption Properties. A number of parameters are used to describe the tion capacity of activated carbon (Sontheimer, Crittenden, and Summers, 1988) The

adsorp-molasses number or decolorizing index is related to the ability of activated carbon to

adsorb large-molecular-weight color bodies from molasses solution, and generallycorrelates well with the ability of the activated carbon to adsorb other large adsor-

bates The iodine number (American Society for Testing Materials, 1988) measures

the amount of iodine that will adsorb under a specified set of conditions, and ally correlates well with the surface area available for small molecules Other num-

gener-bers have been developed for specific applications, such as the carbon tetrachloride activity, the methylene blue number, and the phenol adsorption value The values of

these numbers give useful information about the abilities of various activated bons to adsorb different types of organics However, isotherm data for the specificcompounds to be removed in a given application, if available, are much better indi-cators of performance

car-Two of the more important characteristics of an activated carbon are its pore sizedistribution (discussed previously) and surface area.The manufacturer provides typ-

ical data that usually include the BET surface area This parameter is determined by

measuring the adsorption isotherm for nitrogen gas molecules and then analyzingthe data using the Brunauer-Emmett-Teller (BET) isotherm equation (Adamson,1982) to determine the amount of nitrogen required to form a complete monolayer

of nitrogen molecules on the carbon surface Multiplying the surface area occupiedper nitrogen molecule (0.162 nm2/molecule of N2) by the number of molecules in themonolayer yields the BET surface area Because nitrogen is a small molecule, it can

* This definition is based on ASTM Standard D2854-83 (American Society for Testing Materials, 1988).

It conflicts with ASTM Standard C128-84 for apparent specific gravity, as used in Chapter 8, which does not

enter pores that are unavailable to larger adsorbates As a result, all of the BET face area may not be available for adsorbates in drinking water.

sur-Tabulations of single-solute isotherm constants are very useful when only roughestimates of adsorption capacity are needed to determine whether a more intensiveanalysis of the adsorption process is warranted The Freundlich isotherm constants

of Dobbs and Cohen (1980), as tabulated by Faust and Aly (1983), are reproduced inTable 13.1 for this purpose.* Values from Speth and Miltner (1998) have also beenlisted Additional values can be found in Sontheimer, Crittenden, and Summers

(1988) These data can be used to judge relative adsorption efficiency The K values

of isotherms that have nearly the same values of 1/n show the relative capacity of

adsorption For example, if a GAC column is satisfactorily removing 2-chlorophenol

[K=51 (mg/g)(L/mg)1/n and 1/n=0.41], the removal of compounds with larger

val-ues of K and approximately the same concentration will very likely be better (An

exception might occur if the organic compounds adsorb to particles that pass

through the adsorber.) If the 1/n values are much different, however, the capacity of

activated carbon for each compound of interest should be calculated at the rium concentration of interest using Equation 13.1, because the relative adsorbabil-ity will depend on the equilibrium concentration The use of isotherm values toestimate adsorber life and PAC usage rate will be discussed later

equilib-GAC Contactors

GAC contactors can be classified by the following characteristics: (1) drivingforce—gravity versus pressure; (2) flow direction—downflow versus upflow;(3) configuration—parallel versus series; and (4) position—filter-adsorber versuspostfilter-adsorber

GAC may be used in pressure or gravity contactors Pressure filters enclose theGAC and can be operated over a wide range of flow rates because of the wide vari-ations in pressure drop that can be used An advantage of these filters is that theycan be prefabricated and shipped to the site A disadvantage is that the GAC cannot

be visually observed with ease Gravity contactors are better suited for use whenwide variations in flow rate are not desirable because of the need to remove turbid-ity, when large pressure drops are undesirable because of their impact on operationcosts, and when visual observation is needed to monitor the condition of the GAC.For many systems the decision between pressure or gravity contactors is made onthe basis of cost Medium-size and large systems normally use gravity contactors.Water may be applied to GAC either upflow or downflow, and upflow columnsmay be either packed bed or expanded bed Downflow columns are the most com-mon and seem best suited for drinking water treatment McCarty, Argo, and Rein-hard (1979) found that carbon fines were produced during packed-bed upflowoperation and not during downflow operation The pulsed-bed contactor can also beused to decrease carbon usage rate from that of a single contactor The flow isapplied upward through the column; the spent GAC, a fraction of the total amountpresent, is periodically removed from the bottom of the column and an equalamount of fresh GAC is applied to the top

* Dobbs and Cohen (1980) and Speth and Miltner (1998) should be consulted to determine the type of activated carbon and the experimental conditions that were used The data of Dobbs and Cohen were not determined in a way that would ensure that equilibrium was achieved for all adsorbates, but are suitable to show relative absorbability of compounds and to make rough estimates of activated carbon life If precise

TABLE 13.1 Freundlich Adsorption Isotherm Parameters for Organic Compounds

Compound K (mg/g)(L/mg)1/n 1/n ReferencePCB 14,100 1.03 *Bis(2-ethylhexyl phthalate) 11,300 1.5 †

Heptachlor 9,320 0.92 *Heptachlor epoxide 2,120 0.75 *Butylbenzyl phthalate 1,520 1.26 †

Hexachlorocyclopentadiene 1,400 0.504 *Dichloroacetonitrile 1,300 0.232 *Toxaphene 950 0.74 *Endosulfan sulfate 686 0.81 †

Hexachlorobenzene 450 0.6 †

Pentachlorophenol 443 0.339 *Pentachlorophenol 436 0.34 *Oxamyl 416 0.793 *Anthracene 376 0.7 †

4-Nitrobiphenyl 370 0.27 †

Styrene 334 0.479 *Fluorene 330 0.28 †

γ-BHC (lindane) 285 0.43 *2,4-Dinitrotoluene 284 0.157 *2-Chloronaphthalene 280 0.46 †

Carbofuran 275 0.408 *Phenylmercuric acetate 270 0.44 †

Acifluorofen 236 0.198 *Metolachlor 233 0.125 *

TABLE 13.1 Freundlich Adsorption Isotherm Parameters for Organic Compounds

Bromobenzene 213 0.364 *Dimethylphenylcarbinol 210 0.34 †

Dinoseb 209 0.279 *4-Aminobiphenyl 200 0.26 †

1,2,4-Trichlorobenzene 157 0.31 †

2,4,6-Trichlorophenol 155 0.4 †

β-Naphthylamine 150 0.3 †

Simazine 150 0.227 *2,4-Dinitrotoluene 146 0.31 †

TABLE 13.1 Freundlich Adsorption Isotherm Parameters for Organic Compounds

(Continued)

Compound K (mg/g)(L/mg)1/n 1/n ReferenceAcenaphthylene 115 0.37 †

Methoxychlor 115 0.36 *4-Chlorophenyl phenyl ether 111 0.26 †

Diethyl phthalate 110 0.27 †

Chlorobenzene 101 0.348 *Chlorobenzene 101 0.348 *2-Nitrophenol 99 0.34 †

Dimethyl phthalate 97 0.41 †

Hexachloroethane 97 0.38 †

Toluene 97 0.429 *Bromoform 92 0.655 *Dicamba 91 0.147 *Chloropicrin 88 0.155 *Pichloram 81 0.18 *2,4-Dimethylphenol 78 0.44 †

4-Nitrophenol 76 0.25 †

Acetophenone 74 0.44 †

1,2,3,4-Tetrahydronaphthalene 74 0.81 †

1,2,3-Trichloropropane 74 0.613 *Ethylene thiourea 73 0.669 *Adenine 71 0.38 †

Dibenzo |a,h| anthracene 69 0.75 †

1,1,1,2-Tetrachlorethane 69 0.604 *Nitrobenzene 68 0.43 †

Isophorone 63 0.271 *Methyl isobutyl ketone 61 0.279 *3,4-Benzofluoranthene 57 0.37 †

Trichloroethene 56 0.482 *2,4,5-Trichlorophenoxy acetic acid 55 0.21 *Trichloroacetic acid 52 0.216 *2-Chlorophenol 51 0.41 †

Benzene 50 0.533 *Dibromochloromethane 47 0.636 *5-Bromouracil 44 0.47 †

Dichloroacetic acid 40 0.462 *1,1-Dichloropropene 35 0.374 *Methomyl 35 0.29 *

Benzo |a| pyrene 34 0.44 †

2,4-Dinitrophenol 33 0.61 †

1,1,2-Trichloroethane 33 0.652 *Isophorone 32 0.39 †

1,3-Dichloropropane 28 0.497 *Thymine 27 0.51 †

Chloral hydrate 27 0.051 *5-Chlorouracil 25 0.58 †

Bis(2-Chloroisopropyl) ether 24 0.57 †

Carbon tetrachloride 23 0.594 *

Single-stage contactors are often used for small groundwater systems, but if morethan one contactor is required, lower activated carbon usage rates can be achieved byarranging the contactors either in series or in parallel as shown in Figure 13.13, possiblyyielding a lower-cost system GAC in a single-stage contactor must be removed aboutthe time the MTZ begins to exit the column (Figure 13.11).At this point only a portion

of the activated carbon is saturated at the influent concentration, so the activated bon usage rate may be relatively high.Alternatively, columns may be arranged in series

car-so that the MTZ is entirely contained within the downstream columns after the leadcolumn has been saturated with the influent concentration.When the activated carbon

is replaced in the lead column, the flow is redirected so that it goes through the freshestactivated carbon last.Thus, the activated carbon “moves” countercurrent to the flow of

TABLE 13.1 Freundlich Adsorption Isotherm Parameters for Organic Compounds

(Continued)

Compound K (mg/g)(L/mg)1/n 1/n ReferenceDalapon 23 0.224 *1,2-Dibromoethane 23 0.471 *1,2-Dibromoethene (EDB) 22 0.46 *Endothall 22 0.329 *Bromodichloromethane 22 0.655 *

1,2-Dichloropropane 19 0.597 *Methyl ethyl ketone 19 0.295 *1,1-Dichloroethene 16 0.515 *

1,1,1-Trichloroethane 13 0.531 *Diquat 12 0.242 *

1,1-Dichloroethane 8 0.706 *Cyclohexanone 6.2 0.75 †

Trichlorofluoromethane 5.6 0.24 †

5-Fluorouracil 5.5 1 †

1,2-Dichloroethane 5 5.33 *2-Chlorothethyl vinyl ether 3.9 0.8 †

Methylene chloride 1.6 0.801 *Acrylonitrile 1.4 0.51 †

* Speth and Miltner, 1998.

† Dobbs and Cohen, 1980/Faust and Aly, 1983.

water, and lower activated carbon usagerates are achieved than with single-stagecontactors Series configuration is bestutilized when the effluent criterion is verylow compared to the influent concentra-tion (Wiesner, Rook, and Fiessinger,1987) The increased cost of plumbingcounters the cost benefit of reduced acti-vated carbon usage rate, however, espe-cially when more than two columns must

be used in series

When parallel-flow activated carbonadsorbers are operated in staggeredmode, they can also be used to decreasethe activated carbon usage rate fromthat which is possible with a single-stagecontactor (Westrick and Cohen, 1976;Roberts and Summers, 1982) Becausethe effluent from each of the units isblended, each unit can be operated until

it is producing a water with an effluentconcentration in excess of the treatedwater goal Only the composite flowmust meet the effluent quality goal.Other flow arrangements can be used to produce lower activated carbon usagerates A parallel-series arrangement of gravity filters is used in North Holland(Schultink, 1982), and Sontheimer and Hubele (1987) report that a similar arrange-ment using pressure filters with two layers of GAC was employed at Pforzheim,West Germany Each of the two layers can be backwashed and replaced indepen-dently, and the order of flow through the layers can be reversed A 35 percent loweractivated carbon usage rate for this system for removing halogenated hydrocarbonsfrom groundwater was reported compared to a single-stage system

GAC contactors can also be classified by their position in the treatment train.Thefilter-adsorber employs GAC to remove particles as well as dissolved organic com-pounds These contactors may be constructed simply by removing all or a portion ofthe granular media from a rapid filter and replacing it with GAC Alternatively, anew filter box and underdrain system for the GAC may be designed and con-structed Graese, Snoeyink, and Lee (1987a) discuss these types of filters in detail.The postfilter adsorber is preceded by a granular media filter, and thus has as its onlyobjective the removal of dissolved organic compounds Backwashing of these adsor-bers is unnecessary for particle removal, but if extensive biological growth occurs,backwashing may be required as often as once per week, especially if immediatelypreceded by ozonation (Fiessinger, 1983; Sontheimer, 1983)

PERFORMANCE OF GAC SYSTEMS

Factors Affecting Organic Compound Removal Efficiency

Adsorbate and GAC properties both have important effects on adsorption that havebeen discussed in earlier sections Additional factors that must be considered in thedesign of full-scale systems are presented here

FIGURE 13.13 Adsorber systems.

GAC Particle Size. The effect of particle size on the rate of approach to rium in isotherm determination was discussed previously It has a similar effect onthe rate of adsorption in columns If the rate of adsorption is controlled by intra-particle diffusion, the time to reach equilibrium with a given solution concentration

equilib-in a column approximates that for a batch test With all other factors constant,decreasing particle size will decrease the time required to achieve equilibrium andwill decrease the length of the MTZ in a column Thus, to improve adsorption effi-ciency and to minimize the size of column required, the particle size selected for acontactor should be as small as possible

The rate of head loss buildup caused by particle removal may limit the size ofGAC that can be used in adsorbers The smaller the GAC, the higher the initial headloss and rate of head loss buildup; thus, cost of energy and availability of head have

an important influence on the GAC size selected for a design Additionally, if a adsorber is constructed by replacing media in an existing rapid filter, turbidityremoval efficiency generally increases as the GAC size decreases If the medium istoo small, however, the rate of head loss buildup because of particle accumulationbecomes excessive, and the net water production will be too small for cost-effectiveoperation The filters also become difficult to clean by backwashing

filter-The commercial sizes of GAC are typically characterized by a relatively largeuniformity coefficient of up to 1.9 This large coefficient causes the bed to restrat-ify more easily after backwashing This large uniformity coefficient also requiresthat greater percent expansion of the adsorber be used during backwash in order

to expand the bottom media (Graese, Snoeyink, and Lee, 1987a) Some GAC ters use GAC with a small uniformity coefficient (∼1.3) in deep beds to improvedepth removal of turbidity and to increase net water production (Graese,Snoeyink, and Lee, 1987b) Mixing of the media in these filters undoubtedly ismore than in filters with large uniformity coefficients, so they should not be used

fil-in applications where desorption from the mixed GAC will require early GACreplacement

Common practice is to use 12 ×40 US Standard mesh (1.68 ×0.42 mm) or similaractivated carbon in postfilter adsorbers, because backwashing is rarely required due

to head loss buildup This carbon size also is commonly used in filter-adsorbers when

it is the only filter medium and the filter depth is less than about 75 cm (30 in.).Deeper beds that will be used to remove turbidity commonly employ 8 ×30 USStandard mesh (2.38 ×0.60 mm) or larger activated carbon to promote longer filterruns The option of using custom-sized GAC to obtain a better media design for aparticular application also is available

Contact Time, Bed Depth, and Hydraulic Loading Rate. The most importantGAC adsorber design parameter that affects performance is the contact time, mostcommonly described by the EBCT For a given situation, a critical depth of GAC and

a corresponding minimum EBCT (Equation 13.8) exist that must be exceeded tocontain the MTZ and minimize or eliminate immediate breakthrough As the EBCTincreases, the bed life or service time (expressed in bed volumes of product water tobreakthrough) will increase until a maximum value is reached Correspondingly, theactivated carbon usage rate will decrease to a minimum value For example, Figure13.14 shows that the operating time or service time of a column will increase withincreasing depth, although the increase is not always linear with depth These curvesare commonly called bed depth–service time curves and may be used to determinethe critical depth as shown in Figure 13.14 Figure 13.14 also shows that the percent-age of activated carbon in a column exhausted at MTZ breakthrough increases asdepth or EBCT increases.The mass of organic matter adsorbed per unit mass of acti-vated carbon increases as percent exhaustion increases and, correspondingly, the

number of bed volumes of water that can be processed before MTZ breakthroughalso increases to a maximum value.

Increasing EBCT or bed depth at a constant hydraulic application rate canimpact treatment costs (Kornegay, 1979; Lee et al., 1983; Wiesner, Rook, andFiessinger, 1987) As contactor size increases, fixed costs increase because of thegreater cost of larger systems Operating costs decrease because of decreasing car-bon usage rate and replacement frequency Thus an optimum depth or EBCT can beachieved

Pilot data that show the effect of increasing EBCT on bed life for volatile organicchemicals (VOCs) have been given by Love and Eilers (1982) In general, these data

showed that the carbon usage rate for cis-1,2-dichloroethylene, 1,1,1-trichloroethane,

and carbon tetrachloride decreased substantially as the EBCT was increased from 6

to 12 min, but little change was noted when the EBCT was increased to 18 min mers et al (1997) reported that for three of the four waters examined for TOC andDBP precursor control, no significant increase in the bed volumes treated or carbonusage rate occurred when the EBCT was increased from 10 to 15 to 20 min Collect-ing pilot plant data for a range of EBCT values is important if the lowest activatedcarbon usage for a given application is to be determined

Sum-FIGURE 13.14 Bed depth–service time and percent exhaustion at

breakthrough versus depth.

Other factors must be considered when selecting a hydraulic application rate Asapplication rate increases, the thickness of the hydrodynamic boundary layerdecreases, although this is not a major effect for most applications because film trans-port often plays a relatively small role in affecting rate of uptake of many compoundsfrom aqueous solution Cover and Pieroni (1969) review data that show that adsor-bers with the same EBCT, but with different hydraulic loading rates, give essentiallythe same performance in terms of number of bed volumes processed to breakthrough

provided the depth is considerably greater than the critical depth LMTZ Head loss willincrease with increasing hydraulic rate, so energy costs must be considered Also, ifthe GAC is being used to remove particles as well as dissolved organics, the effect ofapplication rate on removal of turbidity must also be considered

The EBCTs in use today range from a few minutes for some filter-adsorbers(Graese, Snoeyink, and Lee, 1987a) to more than 4 h for the removal of high con-centrations of some specific contaminants (Snoeyink, 1983) Hydraulic applicationrates vary from 1 to 30 m/h (0.4 to 12 gpm/ft2), with a typical value being 7 to 10 m/h(3 to 4 gpm/ft2)

Backwashing. Backwashing of GAC filter-adsorbers is essential to remove solids,

to maintain the desired hydraulic properties of the bed, and possibly to control logical growth, and is often necessary for postfilter adsorbers Backwashing should

bio-be minimized, however, bio-because of its possible effect on adsorption efficiency ing of the bed may take place during backwashing; if it does, GAC with adsorbedmolecules near the top of the bed will move deeper into the bed where desorption ispossible Molecules that are easily reversibly adsorbed, such as carbon tetrachlorideand other VOCs, may be partially desorbed in this new position, leading to a spread-ing out of the MTZ and to early breakthrough (Wiesner, Rook, and Fiessinger,1987) Desorption will not occur if the molecules are irreversibly adsorbed, or if theyare removed by a destructive mechanism, such as biodegradation, instead of adsorp-tion The large uniformity coefficient of most commercial activated carbons pro-motes restratification after backwash, but if the underdrain system does notproperly distribute the washwater, or if the backwash is not carried out in a mannerthat aids restratification, substantial mixing of the activated carbon can occur witheach backwash (Graese, Snoeyink, and Lee, 1987a) Hong and Summers (1994) haveshown that backwashing had little impact on the time to 50 percent breakthroughfor the TOC and THM precursors of four waters For two of the waters, someincrease in TOC after backwashing was detected, but the maximum amount wasonly 0.2 mg/L

Mix-Biological Activity. Biological activity on GAC has several beneficial aspects cific compounds, such as phenol (Chudyk and Snoeyink, 1984), the odor-causing com-

Spe-pounds geosmin and MIB (Namkung and Rittman, 1987; Silvey, 1964), p-nitrophenol

and salicylic acid (DeLaat, Bouanga, and Dore, 1985), ammonia (Bablon,Ventresque,and Ben Aim, 1988), trichlorobenzene (Summers et al., 1989), bromate (Kirisits andSnoeyink, 1999), and probably many more compounds (Rittmann and Huck, 1989),can be removed by biological oxidation rather than adsorption Some evidence forbiological removal of chlorinated benzenes and aromatic hydrocarbons was alsofound in reclamation water (McCarty, Argo, and Reinhard, 1979) Additionally, someportion of the DOC in natural waters can be biologically oxidized on activated car-bon as shown in Figure 13.15 Sontheimer and Hubele (1987) found a small amount

of biological oxidation if the water was not preozonated, but application of 1.1 mg

O3/mg DOC resulted in removal of 35 to 40 percent of the influent DOC by cal oxidation Biodegradable compounds may be removed by microbes, without prior

biologi-adsorption to the GAC, if a biofilm capable of degrading such compounds is oped before they are applied Adsorbable biodegradable compounds may beadsorbed first if the biofilm is not developed when the compounds enter the column,and then desorbed and degraded as the biofilm develops The use of GAC in biofil-ters has recently been reviewed by Servais et al (1998).

devel-Preozonation does not always produce increased amounts of biological tion Glaze et al (1982) studied a GAC system with a 24-min EBCT; one set ofcolumns was preceded by ozonation at 0.5 to 0.6 mg O3/mg TOC and another paral-lel system was not preceded by ozone Typical results showed about 40 to 50 percentTOC removal by GAC for the 20- to 44-wk period of operation in both systems, thusindicating little effect of the ozone About 0.6 to 0.8 mg/L of the approximately

oxida-2 mg/L TOC being removed by each column could be attributed to biological ity It is possible that no significant effect of the ozone was noted because the dose is

activ-on the lower end of the 0.5- to 1.0-mg O3/mg TOC dose recommended for ing biological oxidation (Sontheimer and Hubele, 1987), or because a large fraction

promot-of the organic matter was biodegradable before it reacted with ozone

An important observation made by Glaze et al (1982) at Shreveport was thatbiological activity was high during the summer months when the water temperaturewas in the 25 to 35°C range, but decreased to a relatively insignificant amount aswater temperature dropped to 8 to 12°C during the winter months Thus, biologicaloxidation may vary significantly throughout the year if low temperatures areexpected (Servais et al., 1998)

Sontheimer and Hubele (1987) give typical process parameters for an GAC system for the type of water they studied (Table 13.2) The 100-g DOC/m3-daybiological oxidation compares well with the 75- to 150-g TOC/m3-day found byGlaze et al (1982)

ozone-TOC removal by biologically active GAC systems seems to be a conservativeindicator for DBP precursor removal (Miltner et al., 1996;Wang, Summers, and Milt-ner, 1995) Table 13.3 shows the removal of TOC, DBP precursors as measured bythe formation potential (FP) for THMs and total organic halide (TOX), and assimi-lable organic carbon (AOC), a measure of the biodegradable fraction of NOM, bybiofiltration of ozonated and settled Ohio River water All filters had a total depth

FIGURE 13.15 DOC removal by adsorption and biodegradation

dur-ing GAC filtration of an ozonated humic acid solution (Source:

Sont-heimer and Hubele, 1987.)

of 0.76 m (30 in.), including a bottom layer of 0.20 m (8 in.) of sand, yielding an all EBCT of 9.2 min The steady-state removals reported in the table were takenbetween 5 and 11 mo of operation The GAC columns outperformed the inert mediacolumns, and the micro- and mesoporous GACs seem to perform best in the biolog-ical mode However, Servais et al (1998) report that the macroporous GAC may bebetter suited for cold waters They also report that EBCTs of 5 min were enough toremove the rapidly biodegradable fraction, except for cold waters where longerEBCTs would be required.

over-An important beneficial effect of ozone-GAC systems was noted by van derGaag, Kruithof, and Puijker (1985) when treating coagulated and settled RhineRiver water The mutagenicity of chlorinated effluent from an ozone-GAC systemwas significantly lower than the mutagenicity of chlorinated nonozonated GACeffluent The authors caution that this response may not be the same for all waters.Sontheimer and Hubele (1987) reference similar results from research in Israel.There the ozone-GAC effluent showed a much lower response in cell tissue tests,thus indicating a lower mutagenicity

Biologically active GAC generally does cause the concentration of isms in a GAC column effluent to be higher than in the influent A thorough analy-sis of this is given by Symons et al (1981) They report data from Beaver Falls,Pennsylvania, that show both coliform and standard plate counts were higher inGAC effluent than influent when the water temperature was greater than about

microorgan-10°C, even though 1 to 2 mg/L of chlorine residual was present in the influent to thebed Apparently the GAC reduced the chlorine and allowed the bacteria to regrow.When the water temperature was below about 10°C, no regrowth was noticed Other

data from Philadelphia showed that coliform organisms such as Citrobacter freundii, Enterobacter cloacae, and Klebsiella pneumoniae were found in the GAC filters In

all cases, postdisinfection produced water meeting United States EnvironmentalProtection Agency regulations

TABLE 13.2 Process Parameters for Activated Carbon

Following Ozone

Parameter ValueOzone dosage 0.5 to 1.0 g ozone/g DOC

Biological degradation ∼100 g DOC /(m3-day)

Oxygen demand for DOC oxidation ∼200 g oxygen/(m3-day)

EBCT 15 to 30 min

Source: Sontheimer and Hubele, 1987.

TABLE 13.3 Impact of Media Type on Performance—Ozonated/Settled Ohio River Water

Removal (%)Anthracite— GAC— GAC— GAC—Parameter sand Sand microporous mesoporous macroporousTOC 16 20 29 27 21AOC-NOX 39 43 51 47 42THM FP 23 23 40 34 27TOX FP 28 25 52 44 31

Even though bacteria grow readily in GAC filters, and high microorganism countsare observed in effluents from these processes, the disinfectant demand to achievemicroorganism kill is much reduced by GAC filtration An exception to this occurs ifactivated carbon particles penetrate the underdrain and provide a habitat for microbesthat protects them from being killed by disinfectant (LeChevallier et al., 1984).Zooplankton can grow in GAC filters if the filters are biologically active Organ-

isms such as oligochaetes and rotatoria have been reported to increase as water was

processed through GAC during the summer months at Rotterdam and North land (van der Kooij, 1983), probably because they use bacteria for food Similarobservations have been made in West Germany (Sontheimer, 1983) and France(Fiessinger, 1983) Sontheimer and Fiessinger recommend backwashing with airscour once every five days or so to wash the eggs of these organisms out of the adsor-ber before they can hatch In contrast, microscreening is used to remove the organ-isms from the GAC-filtered water at North Holland

Hol-Control of Microbial Growth. Biologically active carbon must be controlled toavoid undesirable effects Anaerobic conditions may develop, with attendant odorproblems, if the system is not kept aerobic This may happen if large concentrations

of ammonia enter the filter (each mg/L of NH3requires about 3.8 mg/L of dissolvedoxygen if it is converted to NO3 −), if insufficient dissolved oxygen is in the water, or

if the bed is allowed to stand idle for a period of time

Control is also possible through proper design Figure 13.16 shows the distribution

of microorganisms in a biologically active GAC filter that is treating nonchlorinatedwater (Sontheimer, Crittenden, and Summers, 1988; Topalian, 1987) Much largernumbers of organisms are on the activated carbon at the entrance to the bed than atgreater depths This distribution is consistent with larger amounts of adsorbed com-pounds in the upper level (Sontheimer and Hubele, 1987) Growth in the upper part

of a deep bed has the opportunity to be removed deeper within the bed when itsloughs off van der Kooij (1983) noted that whereas increasing the EBCT by increas-ing the bed depth can bring about sharp reductions in numbers of organisms in GACfiltrate, increasing EBCT by decreasing flow rate did not decrease, but sometimesincreased, the plate count Available data on the effect of backwashing on effluentorganism concentration are inconclusive (van der Kooij, 1983)

FIGURE 13.16 Microbiological parameters in a

GAC filter (Source: Sontheimer and Hubele, 1987.)

Application of chlorine to GAC adsorbers does not prevent growth, andincreases the concentration of adsorbed chloroorganics The potential exists forchlorine to make the activated carbon become more friable and break up more eas-ily, especially during backwash, because chlorine destroys some of the activatedcarbon when it is reduced (Snoeyink and Suidan, 1975) Also, the potential existsfor formation of unique organic compounds through the catalytic action of the acti-vated carbon surface, as shown later Thus, application of chlorine to GAC filters isnot recommended.

Pretreatment for GAC Systems. Pretreatment can have a significant impact on theperformance of activated carbon systems The influent concentration of organicsmay be lowered, the species of compounds may be changed, thus changing adsorba-bility and biodegradability, and the inorganic composition may be changed in a man-ner that affects absorbability and the tendency of the activated carbon to becomefouled The removal of NOM by coagulation, sedimentation, and filtration reducesthe quantity of organics that must be removed by adsorption Lower-cost operation

of GAC systems for TOC and DBP precursor removal may then be possible (seeChapter 10 for a discussion of organic compound removal by coagulation)

Summers et al (1994) and Hooper et al (1996a), as shown in Figure 13.17, havesummarized the impact of the initial TOC concentration TOC0on the run time to a

50 percent TOC breakthrough, measured as bed volumes BV50for a wide range ofsource waters The relationship can be expressed as

BV50=18,000/TOC0Twenty-eight case studies of GAC bench, pilot, and full-scale contactors from 21different source waters were evaluated All systems utilized bituminous-coal-basedGAC, and the influent pH was between 7 and 8 for the river, lake, and groundwaterexamined The relationship was verified by five GAC runs of the same isolatedNOM, diluted to different concentrations (hollow symbols in Figure 13.17) Part ofthe variability of the data around the regression line is likely due to differences inthe adsorbability of the NOM caused by pretreatment and differences betweensources

FIGURE 13.17 Correlation between influent TOC concentration and bed

volumes to 50 percent TOC breakthrough (Source: Hooper et al., 1996a.)

FIGURE 13.18 Effect of pH on TCP, DCP, and soil vic acid (FA) adsorption (TCP and DCP data taken from Murin and Snoeyink, 1979; FA data, expressed in moles C, taken from McCreary and Snoeyink, 1980.)

ful-Adsorption of organic acids and bases by GAC is generally affected by solution

pH, so pretreatment steps that affect pH have an important effect on adsorption Ingeneral, both the undissociated and ionized forms of an adsorbate can be adsorbed

on GAC, with the undissociated form being more strongly adsorbed The ionic formhas a much higher affinity for polar water molecules and thus tends to remain insolution Typical data are shown in Figure 13.18 for three weak organic acids: 2,4,6-trichlorophenol, 2,4-dichlorophenol, and soil fulvic acid (McCreary and Snoeyink,1980; Murin and Snoeyink, 1979) Different types of humic substances may haveboth different capacities and capacity dependencies on pH

Several researchers have shown the impact of influent pH on the adsorption ofTOC Unfortunately, some of the work has been done with different initial TOC con-centrations, and the increased performance attributed to low pH may be because ofthe lower TOC0 (Figure 13.17) A relationship between the relative adsorptioncapacity for TOC and pH is shown in Figure 13.19 for 13 different source waters and

a bituminous-coal-based GAC (Hooper, Summers, and Hong, 1996) Within the pHrange shown, a decrease in the pH of one unit yielded a 6 percent increase in adsorp-tion capacity

Several investigators have reported better GAC performance for TOC controlafter coagulation or after increasing the coagulant dose to achieve enhanced coagu-lation Hooper and coworkers (Hooper, Summers, and Hong, 1996; Hooper et al.,1996a,b) have shown that the increase in GAC run time after enhanced coagulationcan be attributed to the lower pH and lower initial TOC concentration associatedwith the coagulated water

Reactions of chlorine, or other oxidative pretreatment chemicals such as oxygen,ozone, chlorine dioxide, and permanganate, with GAC or with organic compounds

in aqueous solution or on the GAC surface can alter the adsorption performance.For example, ozone can react with humic substances to produce more polar inter-

mediates that are less adsorbable on GAC (Chen, Snoeyink, and Fiessinger, 1987)but usually more biodegradable (Sontheimer and Hubele, 1987) If TOC is morebiodegradable, increased removals by microbiological activity in a GAC contactorare expected.

However, if biological treatment is not effective, then the weakly adsorbing pounds can have a negative impact on the overall GAC performance Solarik et al.(1996) systematically evaluated the impact of ozonation and biotreatment on subse-quent GAC performance for five waters, as illustrated in Figure 13.20 for Ohio Riverwater.They showed that ozonation and biotreatment decreased the humic and inter-mediate molecular size fractions, which are the most strongly adsorbing fractions.They found that the early part of the breakthrough was dominated by the relativeincrease in the weakly adsorbing fraction, which in some cases led to earlier break-through, while the overall lower influent TOC concentration dominated the latterportion of the breakthrough curve and in some cases led to longer run times

com-FIGURE 13.19 Impact of pH on percent DOC adsorbed for several

waters (Source: Hooper, Summers, and Hong, 1996.)

FIGURE 13.20 Effect of pretreatment on TOC breakthrough for

Ohio River water (Source: Solarik et al., 1996.)

Chlorine-containing disinfectants (HOCl, ClO2, or NH2Cl) may react both withactivated carbon and adsorbed compounds Unusual products not characteristic ofsolution reactions may be formed when activated carbon is present For example, theHOCl reaction with adsorbed 2,4-dichlorophenol (2,4-DCP) resulted in a series ofhydroxylated PCBs at HOCl concentrations normally encountered in drinking watertreatment practice (see Table 13.4).A similar product mixture was also obtained whenthe GAC was first treated with HOCl and then 2,4-DCP was adsorbed The HOCl-activated carbon surface caused similar compounds to form as adsorption took place.Furthermore, some of these products may desorb from the activated carbon column.Additional data are given on this effect by Voudrias, Larson, and Snoeyink (1985).Thereaction products that have been found to date have been formed only in laboratorysystems Further research is needed to show whether they will also form in field instal-lations in the presence of humic substances Because such compounds might form, andbecause their health effects are unknown, the application of chlorine-containing disin-fectants to GAC adsorbers needs to be eliminated where possible.

Vidic and Suidan (1991) showed that dissolved oxygen also reacted withadsorbed compounds such as phenols, converting them to products that allowedmore of the target adsorbate to be removed from solution The capacity for thesecompounds was thus much greater if oxygen was present than if the solution wasfree of oxygen

Pretreatment to prevent fouling of GAC is also important Application of waterthat is supersaturated with salts such as calcium carbonate will lead to blockage ofthe activated carbon pores and possibly to complete coverage of the particle Ironand manganese precipitates may also interfere with adsorption If ammonia concen-

Source: Reprinted with permission from E A.

Voudrias, R A Larson, and V L Snoeyink “Effects of activated carbon on the reactions of free chlorine with

phenols.” Environmental Science and Technology,

TABLE 13.4 Reaction Products from 2,4-Dichlorophenol-GAC Reaction

HOCl-tration is very high, pretreatment for ammonia removal may also be necessary toprevent depletion of dissolved oxygen in biologically active beds.

Reactions of Inorganic Compounds with Activated Carbon

Activated carbon in water treatment may inadvertently contact oxidants such asoxygen, aqueous chlorine, chlorine dioxide, and permanganate and react with them.Virgin activated carbon has been shown by Prober, Pyeha, and Kidon (1975) to reactwith 10 to 40 mg aqueous O2per gram of carbon over a time span of 1700 h, andChudyk and Snoeyink (1981) found that 3 to 4 mg O2per gram of carbon reactedover a time span of 130 h Some of this oxygen is converted to surface oxides(Prober, Pyeha, and Kidon, 1975)

Free Chlorine-Activated Carbon Reactions. The well-known reactions of HOCland OCl−with activated carbon are as follows:

HOCl +C* →C*O +H++Cl− (13.10)

where C* and C*O represent the carbon surface and a surface oxide, respectively.These reactions proceed rapidly, with the one in Equation 13.10 (pH <7.5) beingfaster than that in Equation 13.11 (pH >7.5) (Suidan, Snoeyink, and Schmitz, 1976,1977a,b) Reactions of free chlorine with activated carbon will result in the produc-tion of organic by-products The TOX on the surface increases as extent of reactionincreases, and some of these compounds may be found in the column effluent if thereaction proceeds for a very long time (Dielmann, 1981)

Combined Chlorine-Activated Carbon Reactions. Bauer and Snoeyink (1973)hypothesized that the following reactions were taking place between monochlor-amine, NH2Cl, and activated carbon:

NH2Cl +H2O +C* →NH3(aq) +C*O +H++Cl− (13.12)

2 NH2Cl +C*O →N2(g) +2H++2Cl−+H2O +C* (13.13)Initially, all the NH2Cl was converted to NH3and Cl−in accordance with Equation13.12, but after a period of reaction, some of the NH2Cl was converted to N2(g) andHCl in accordance with Equation 13.13 The rate of reaction was much slower thanthe reactions of either free chlorine or dichloramine with activated carbon (Kim andSnoeyink, 1980) Additional studies of this reaction were made by Komorita andSnoeyink (1985), who showed that the rate of reaction was high initially and thenreached a plateau value The higher initial rate is useful in design of some GACdechlorination systems

Dichloramine (NHCl2) reacts very rapidly with activated carbon according to thefollowing reaction (Bauer and Snoeyink, 1973; Kim, Snoeyink, and Schmitz, 1978):

2NHCl2+H2O +C* →N2(g) +C*O +4H++4Cl− (13.14)When excess ammonia was present, Kim, Snoeyink, and Schmitz (1978) found evi-dence of the parallel reaction,

NH++3NHCl →2N(g) +7H++6Cl− (13.15)

Bromate, Chlorine Dioxide, Chlorite, and Chlorate-Activated Carbon Reactions.

Other halogen oxides react with activated carbon For example, Miller, Snoeyink, andHarrell (1995) showed that BrO3 −was converted to Br−by virgin activated carbon.ClO2reacts rapidly with activated carbon, but the nature of the reaction changes as

pH changes At pH 3.5, Cl−, ClO2 −, and ClO3 −are found in the effluent of a columnreceiving only ClO2, but Cl−is the predominant product At pH 7.9, the same endproducts are formed, but ClO2 −is now the predominant species (Chen, Larson, andSnoeyink, 1982) When ClO2 −solutions at pH 7 were applied to virgin GAC, ClO2 −readily reacted, presumably in accordance with the reaction (Voudrias et al., 1983)

The reaction capacity of the fresh carbon was saturated after 80 to 90 mg ClO2 −reacted per gram of GAC The chlorate ion was not reduced by activated carbon, butwas adsorbed to a slight extent (∼0.03 mg ClO3 −/g), presumably by an ion exchangemechanism (Dielmann, 1981)

Removal of Other Inorganic Ions. Huang (1978) reviewed the removal of ganic ions by activated carbon A number of ions can be removed from water byGAC, but the capacity for most substances is quite low The gold-cyanide complex,for example, is adsorbed on GAC in a widely used method for recovering gold in themining industry Huang reviewed data that show some removal of cadmium (II) athigher pH, and that removal can be increased slightly by complexing before adsorp-tion with chelating agents The removal of chromium involves adsorption of Cr(III)

inor-or Cr(VI), and under some conditions Cr(VI) is chemically reduced to Cr(III) by theactivated carbon Mercury adsorption is best at low pH From 0.1 to 0.5 mmole Cu/g

at equilibrium concentrations of 8 mmol/L have also been observed to adsorb vated carbon will also catalyze the oxidation of Fe(II) to Fe(III)

Acti-Adsorption Efficiency of Full-Scale Systems

Taste and Odor Removal. Many types of taste and odor problems are tered in drinking water Troublesome compounds may result from biological growth

encoun-or industrial activities They may be produced in the water supply, in the water ment plant from reactions with treatment chemicals, in distribution systems, and inconsumers’ plumbing systems (American Water Works Association, 1987) Acti-vated carbon—both PAC and GAC—has an excellent history of success in removingtaste and odor compounds from raw water

treat-GAC filter-adsorber systems are reported to effectively remove odor fromsource water for typically one to five years (Graese, Snoeyink, and Lee, 1987b) Thebed life is dependent upon the intensity and frequency of appearance of taste andodor compounds, the presence of organics that compete for adsorption sites, and theconcentration of these compounds that is acceptable in the treated water Case his-tory information is difficult to apply at different utilities because the intensity andfrequency of appearance usually is not well documented and the acceptable level oftaste and odor varies from community to community

Regina, Saskatchewan, is an interesting example of the effect of high-intensitytastes and odors on GAC performance (Gammie and Giesbrecht, 1986; Snoeyink andKnappe, 1994) The water treatment plant has postfilter-adsorbers with an EBCT of

15 min and a bed depth of 10 ft The GAC was reactivated once per year The influentwater had a DOC concentration of 2 to 3 mg/L and typical threshold odor numbers(TONs) of 5 to 15 with spikes of 40 to 60 The effluent goal was a TON of 1

Figure 13.21a depicts TON breakthrough curves for different depths within the

GAC adsorber as a function of the number of bed volumes treated The figure showsthat the optimal filter depth was 5 ft At greater filter depths, increasing the beddepth reduced the bed volumes treated before breakthrough This observation isindicative of preloading* by NOM, which reduces the accessible adsorption sites fortaste- and odor-causing substances By 8000 bed volumes, 50 percent of the DOChad broken through Also, a bed depth of 3 ft appeared to be too short to effectivelyreduce TON

Figure 13.21b shows the progression of the TON mass transfer zone through the

GAC bed After 18 days of operation, the TON reached a value of 3 at a GAC depth

of 1 ft, and after 60 days the same TON level was found at a bed depth of 3 ft After

130 days of operation (13,000 bed volumes), the entire GAC bed had become fective for reaching the effluent TON goal

inef-* Preloading means the adsorption of NOM or similar organic matter in the lower section of the

adsor-ber, thereby reducing the adsorption capacity for the target compound, before the target compound reaches

FIGURE 13.21 TON adsorption at Regina (Source: Snoeyink

and Knappe, 1994.)

(a) TON Breakthrough Curves

(b) TON Mass Transfer Zone

Some useful observations have been made using data from full-scale GAC tems Background organic matter, for example, usually breaks through much earlierthan the taste and odor compounds (Love et al., 1973; Robeck, 1975), as can be

sys-observed in Figure 13.21a.

The experience at Stockton East Water District, Stockton, CA, showed that tors other than adsorption may complicate the removal of MIB and geosmin(Thomas, 1986) GAC filters were operated over a period of approximately 2 yrwhen MIB and geosmin in the adsorber influent were often in the 5- to 20-ng/Lrange Extraction of GAC taken from cores of the filter after about 2 yr of operationshowed no measurable MIB, and geosmin levels of about 0.1 to 0.2 µg/g GAC in theupper part of the filter (detection limits were 0.01 µg/g) These measurements wereconsistent with the concept that MIB, and possibly some geosmin, were removed bybiodegradation, because both compounds are biodegradable (Namkung andRittmann, 1987; Silvey, 1964)

fac-One hypothesis for an MIB and geosmin removal mechanism is that adsorptioncapacity is needed to prevent these molecules from passing through the GAC adsor-ber After the initial accumulation on the GAC, a biofilm develops that is capable ofbiodegrading these compounds They then desorb and diffuse to the biofilm, wherethey are biologically oxidized The ability of GAC to remove odor decreases overtime because adsorption sites are gradually taken up by natural organics and thusare not available for MIB and geosmin GAC replacement is necessary whenadsorption capacity has been used up by these competitors

GAC adsorbers with an 11-min EBCT were used to control a sulfide odor lem at the Goleta Water District in California (Lawrence, 1968) The odor wasattributed to hydrogen polysulfides in the groundwater that were not removed byaeration and that GAC was only partially effective in removing (Monsitz andAuinesworth, 1970) Influent TON varied from 6 to 1000, while effluent TON rangedfrom odor free to 35 (Love et al., 1973) These beds were in service for 2 yr beforethe activated carbon was replaced

prob-Total Organic Carbon and DBP Precursors. The type and adsorbability of NOM

in water vary widely from location to location Sufficient differences in adsorbabilityexist, with respect to the quantity of TOC adsorbed and the competitive effect of theTOC on adsorption of trace organic compounds, that adsorption tests should bedone on the water in question if the results are needed for design However, Figure13.17 can be used to provide estimates of the range of adsorbability for convention-ally treated waters

Roberts and Summers (1982) presented an excellent summary of full-scale plantperformance for TOC removal They evaluated removals from 47 different plants,including some wastewater reclamation plants The ranges of design conditions aregiven in Table 13.5, and typical breakthrough curves are given in Figure 13.22 The

TABLE 13.5 Design Conditions for GAC Adsorbers

Parameter Median Range Typical range

breakthrough curves reach an effluent concentration plateau about 10 to 25 percentbelow the influent concentration Roberts and Summers found that the amount oforganic matter remaining in the effluent immediately after start-up decreased as theEBCT increased to about 20 min, and that the time to reach a steady-state effluentconcentration increased as the EBCT increased.

Graese, Snoeyink, and Lee (1987a) summarized TOC removal data for a criterion

of 50 percent removal and showed that adsorbers with EBCT values of less than 10min had a life of less than 30 days, and that service time increased as EBCTincreased To obtain the lowest-cost adsorption system for TOC removal, EBCTs inthe range of 10 to 20 min should be closely examined

The performance for GAC removal of DBP precursors usually parallels that forthe removal of TOC (Hooper et al., 1996a; Lykins, Clark, and Adams, 1988) Symons

et al (1981) have summarized much of the research prior to 1981 Hooper et al.(1996a), Solarik et al (1995a,b), and Summers et al (1994) assessed the use of TOCand UV absorbance as indicators of DBP precursor breakthrough for seven sourcewaters and the following DBPs: THMs, haloacetic acids (HAAs), TOX, and chloralhydrate (CH) The precursors were assessed using the uniform formation conditions(UFC) test (Hooper et al., 1996a) In all but one water, TOC was a good conserva-tive surrogate in that it broke through prior to the DPB precursors, as illustrated inFigure 13.23 In one case, THM precursors broke through a filter adsorber with anEBCT of 6.3 min prior to TOC Organics measured at a UV absorbance wavelength

of 254 nm tended to break through after the DBP precursor TOC removal can varyseasonally because the adsorbability and initial concentration may change with sea-son (Solarik et al., 1995b)

The bromide concentration of the water has an important bearing on the use ofGAC for DBP control Symons et al (1981) observed that formation of brominatedTHMs occurs more rapidly in GAC adsorber product water Graveland, Kruithof,and Nuhn (1981) made a similar observation, and noted a shift to the formation ofmore highly brominated forms of THMs in GAC effluent This stems from the rapidformation of bromine-substituted THMs compared to chloroform when precursorconcentration is low during the first part of the breakthrough curve, because theDBP formation reactions are precursor limited (Summers et al., 1993)

FIGURE 13.22 Representative TOC breakthrough curves (Source: Roberts and Summers, 1982.)

A pilot study that evaluated the removal of organic carbon by GAC and BAC atthe pilot scale was conducted in Newport News, VA (Snoeyink and Knappe, 1994).The design of the pilot test consisted of two columns in series for both the GAC andthe BAC tests, with each column having an EBCT of 15 min The influent to both theGAC and the BAC columns consisted of coagulated, settled, and filtered water; inaddition, the BAC influent was ozonated The TOC concentration of the influentwas 3 mg/L, which gave a distribution system THMFP of 60 µg/L before ozonation.The duration of the study was 107 days for the GAC columns and 83 days for theBAC columns Figure 13.24 depicts the THMFP breakthrough curves at differentEBCTs In both cases, EBCT had little impact on the breakthrough curves, with theexception of the GAC breakthrough curve for an EBCT of 5 min (The 5-min curvewas considered unreliable because of the difficulty in maintaining a constant depth

of GAC above the first sampling port.) A comparison of GAC and BAC curves inFigure 13.24 shows that both the GAC and the BAC treatment trains resulted inbreakthrough curves that were beginning to plateau at a dimensionless THMFPconcentration of about 0.7, an indication that biological removal of organic carbonoccurred both for the ozonated and the nonozonated influent Additional run time

is needed to fully establish the effect of ozone on the plateau However, the shape ofthe breakthrough curves suggests that biological activity developed earlier in theBAC columns than in the GAC columns

These results highlight some important issues that must be considered when ing with biologically active filters It must be determined whether the plateau regionmeets the treatment objective and whether the plateau region is a function of theozone dose that was applied to the filter influent and EBCT If the plateau regionmeets the treatment objective, very long run times can be achieved with BAC filters.Long pilot test runs are required to establish the plateau value as a function ofEBCT The life of the filter as a function of ozone dose and EBCT also needs to beexamined if the treatment objective is exceeded before the plateau is reached Asshown in the Newport News pilot study, the life of the filter would have been greaterfor the BAC than for the GAC for a treatment objective that was reached before theplateau, because the BAC breakthrough curve was not as steep

deal-FIGURE 13.23 Comparison of RSSCT, UV 254 , and UFC-DBP breakthrough

for Salt River Project water (Source: Hooper et al., 1996.)