ACTIVATED CARBON ADSORPTION OF DIBROMOCHLOROPROPANE
Modeling of Adsorber Performance Under
Conditions of Water Treatment
By
Douglas D. Endicott
June 1988
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Drinking Mater
Technical Support Division
Cincinnati, OH 45268

-------
TABLE OF CONTENTS
Page
INTRODUCTION 		J
Dibromochloropropane 		2
Modeling of Target Contaminant Adsorption 		3
MATERIALS AND METHODS 			J
Mathematical Adsorption Models 		'
Experimental Apparatus 		7
Adsorption Isotherms 		'
Kinetic Experiments 		®
RESULTS AND DISCUSSION 		9
Adsorption Isotherms 		9
DCBR 		10
Short-Bed Adsorber 		J*
Dynamic Verification Column 		*6
Adsorption MTP Correlations 		18
MODELING APPLICATIONS 		19
CONCLUSIONS 		23
REFERENCES 		25
TABLES 	 30-34
FIGURES
APPENDIX - EXPERIMENTAL DATA

-------
INTRODUCTION
The Safe Drinking Water Act(47) and the subsequent Amendments of
1986(48) provide the U.S. Environmental Protection Agency (EPA) the
mandate to regulate contaminants in drinking water, in order to protect
the health of persons served by public water systems. Congress has
directed EPA to establish maximum contaminant limits (MCLs) for
microbiological, inorganic, radionuclide, synthetic organic and
disinfectant by-product parameters of health concern in drinking water.
EPA must also identify best available technology (BAT) treatment
techniques for removing these contaminants from drinking water supplies.
The ability of activated carbon to efficiently remove a broad
spectrum of organic contaminants from water(63), even at low
concentrations, has placed activated carbon adsorption among the BAT
candidates for several categories of drinking water contaminants.
Activated carbon adsorption is considered to be an advanced treatment
technology, which has seen only limited application for removing drinking
water contaminants of health concern. The development and application of
activated carbon adsorption for water treatment may be expected to play an
important part in achieving the goals of the SDWA.
EPA has investigated the application of activated carbon for removing
organic contaminants from drinking water for the last decade. In 1978,
EPA proposed a treatment requirement regulation, requiring the
installation of activated carbon by water utilities using surface water
supplies considered to be vulnerable to possible contamination by
synthetic organic chemicals(17). The proposal was withdrawn in 1981, due
to the widespread opposition of water treatment professionals and the
water supply industry(56). In subsequent regulations and proposed
regulations under the SDWA and its 1986 amendments, EPA has named
activated carbon adsorption as BAT(14) for volatile organic compounds
(VOCs), SOCs and disinfection by-products (DBPs). Whether or not
activated carbon adsorption ever achieves widespread application as a
drinking water treatment technique will depend upon a number of uncertain
factors. These include the organic contaminant values in relation to the
concentration and frequency of their occurrence in public water supplies.
Two approaches have been pursued by EPA in determining the
feasibility of activated carbon adsorption as a drinking water treatment
technique. The first has been the construction and operation of
pilot-scale adsorbers at several water treatment plants, to evaluate
long-term operation under field conditions(34). Jefferson Parish,
Louisiana; Miami, Florida; and Cincinnati, Ohio, were primary sitesof
this work. The second approach has been the development of adsorption
data in small-scale laboratory experiments, fora much larger list of
organic contaminants. A peer review of adsorption research conducted by
the EPA Drinking Water Research Division, bringing together many of the
prominent water treatment researchers and professionals, recommended that
mathematical models of the adsorption process be used in conjunction with
the data-gathering efforts in order to more rapidly develop treatability

-------
2
information(58). The role of modeling in expanding the data base of
information for organic contaminant adsorption from drinking water became
the impetus for the work described herein.
The purpose of this study was threefold:
1.	Develop the capability to conduct adsorption experiments and
perform mathematical modeling, within the Technical Support
Division.
2.	Develop activated carbon adsorption treatability data for
organic drinking water contaminants of regulatory interest to
EPA. The work presented in this report focused upon the
adsorption of dibromochloropropane, a contaminant to be
regulated as a synthetic organic compound.
3.	Examine the applications and limitations of adsorption models
for regulatory and engineering decision making.
Pi bromoch1oropropane
l,2-dibromo-3-chloropropane (DBCP) is a soil fumigant and nematocide,
as well as an intermediate in organic synthesis(59). It was registered as
a pesticide in 1948, and has been widely used in agricultural regions of
this country. Between 1977 and 1979, the use of DBCP was restricted to
pineapple fields in Hawaii. All uses in the U.S. were cancelled in 1979.
DBCP has been detected in ground water in six states - Hawaii, California,
Arizona, Maryland and Wisconsin - at concentrations of 0.02 to 20
ug/L(33). Sources of ground water contamination include industrial spills
and discharges as well as agricultural application. Several physical-
chemical properties of DBCP are presented in Table 1. Due to the
compound's moderate volatility and solubility, resistance to biodegra-
dation, and hydrolytic stability, DBCP is quite persistent in the
subsurface environment(1).
The human health effects of exposure to DBCP in drinking water are
uncertain(18). DBCP has been shown to cause sterility in male chemical
workers exposed to high industrial concentrations. Association between
forms of cancer and DBCP have been found in rats and mice. Human cancers
resulting from DBCP exposure have not been documented, and the IARC has
classified the compound as having inadequate evidence for carcinogenicity
in humans. The Hawaii State Health Department has begun monitoring acute
and chronic effects of DBCP upon a population exposed to the contaminant
in drinking water(28).
Based upon the occurrence of DBCP in ground water constituting
drinking water supplies, and the possible adverse health effects of
exposure, EPA has proposed to regulate DBCP as a synthetic organic
compound under the National Primary Drinking Water Regulations(18,19). A
maximum contaminant level goal (MCLG) for DBCP has been set at a concen-
tration of zero. The enforceable drinking water limit, the maximum
contaminant level (MCL), has been proposed at a concentration of 0.2 ug/L.

-------
3
Two technologies, packed-tower aeration and GAC adsorption, are
considered by EPA to be best available treatment (BAT) for removing
synthetic organic chemicals from drinking water supplies. For DBCP,
laboratory and field studies have been conducted for both technologies,
providing a preliminary indication of their effectiveness. The removal of
DBCP from water by packed-tower aeration was evaluated in several field
study(12,50). Results indicated that high air-to-water ratios (on the
order of several hundred) would be necessary to achieve removal
efficiencies of 90% or more.
Data are available from several studies of GAC adsorption of DBCP.
The performance of small, in-home GAC adsorbers removing DBCP from ground
water was monitored in one study(23). At several installations, treating
water with DBCP concentrations of 9.2 to 16 ug/L, a treatment objective of
1 ug/L was met at loading rates of 2.2 to 4.4 mg DBCP removed per gram
activated carbon. Another study(39) reported on the performance of a
full-scale GAC system treating a ground water contaminated by DBCP (77-184
ug/L), EDB (8-19 ug/L), and numerous other SOCs. No breakthrough was
detected in 12 months of operation, equivalent to an activated carbon
usage rate of greater than 2.1 mg/gm. Pilot studies(2) for the same
project, conducted in water containing significantly higher DBCP
concentrations (530-1380 ug/L), obtained a loading rate of approximately 1
mg/gm. Several other laboratory and pilot studies are reported in the
literature(44,20). EPA's Drinking Water Research Division has developed
adsorption data for DBCP, including results from bottle-point isotherm and
dynamic minicolumn experiments(41). These data demonstrate a high
affinity of DBCP for activated carbon, which is diminished in a surface
water matrix (Ohio River water). It may be concluded from the available
information that GAC adsorption is effective for removing DBCP from water,
although the performance of the process is somewhat variable.
Deriving information from field data, in particular, is difficult for
a process such as adsorption where a substantial number of uncontrolled
variables may affect the observed performance. In the next section,
mathematical models are introduced as a tool for evaluating adsorption
performance.
Modeling of Target Contaminant Adsorption
Mathematical adsorption models were developed as research tools, to
aid in interpreting and understanding the adsorption process(7). They have
been successful in simulating the performance of fairly simple
experimental adsorber systems. This success has led to interest in using
adsorption models for engineering applications.
The design of a GAC adsorption system that is economical and
effective requires that the process be optimized with regards to EBCT and
flow configuration of adsorbers, selection of activated carbon,
pretreatment, and backwashing. Determining the optimum choice of design
parameters requires the engineer to bring together consideration of
process aspects, including adsorption equilibrium and kinetics, multi-
component interactions, and reactor design. How optimum design parameters

-------
4
are influenced by process aspects depends upon the contaminant(s) present,
treatment objectives, and other site-specific factors including
characteristics of the water. Pilot adsorber studies are usually used to
select the full-scale design, by a process of varying pilot design
parameters and observing the resulting performance(21). However, the use
of mathematical models of the adsorption process for design has been
advocated(61), and interest in their practical applications is growing.
Predictive models offer the powerful attraction of allowing the engineer
to simulate process performance for a wide range of design parameters,
rapidly and inexpensively. Experience to date indicates that models may
not be adequate substitutes for pilot studies, however, they are useful
for designing more efficient pilot experiments. Furthermore, they may be
used as scale-up(ll) and optimization tools in conjunction with pilot
studies.
Adsorption models are also useful for regulatory purposes, to aid in
the process of rule-making and standard setting. Adsorption is a
treatment candidate for a large number of significant drinking water
contaminants: VOCs, SOCs, DBPs, and DBPPs (disinfection by-product
precursors). Federal and State regulators are faced with rapidly
establishing allowable limits for these contaminants in drinking water to
protect public health; as a part of this decision-making they must
determine from limited data what performance can be achieved by available
treatment technologies. Properly applied, models can assist in this
determination. They may be calibrated with existing data, and then used
to forecast performance for a variety of treatment conditions. The
accuracy of model predictions performed in this way is limited by the
inability to incorporate site-specific factors influencing performance in
the calibration process. This should not compromise the usefulness of
models for regulatory evaluation of adsorption performance, so long as
they are properly applied and calibrated to appropriate data sets(58).
The practical application of adsorption models is limited by a lack
of model testing under full-scale, water treatment conditions. Only a few
field studies of adsorption have incorporated mathematical models.
Nausau, Wisconsin, was the site of the first comprehensive, model-based
investigation of adsorption under drinking water treatment conditions.
Crittenden and coworkers(43) studied the removal of a multicomponent
mixture of VOCs from ground water. The homogeneous surface diffusion
(HSD) and pore-surface diffusion (PSD) models proved capable of predicting
breakthrough dynamics of the weakly adsorbing VOCs, such as
cis-l,2-dichloroethylene, for which the best set of data was
available(lO). Other field investigations of adsorption modeling, in
Germany, indicate some limitations in modeling more strongly-adsorbed
S0Cs(66). Both groups of investigators have begun to look at the role
played by natural organic natter (NOM), present in the water, upon
adsorption of SOCs by GAC. Other aspects of adsorption performance under
study include the impact of humic-binding of hydrophobic compounds upon
their subsequent adsorbability; understanding, parameterizing, and
modeling the role of microorganisms in removing/transforming target
contaminants in GAC beds(57,52); and the effect of backwashing and
activated carbon selection upon performance.

-------
5
The fact that natural organic natter present in water affects the
adsorption of target contaminants is widely acknowledged(35), but the
phenomenon is not well understood. Several investigators have studied
the adsorption of NOM, principally humic and fulvic acids(30,36,62). NOM
Is relatively weakly adsorbed by GAC, with very slow adsorption
kinetics(54) - NOM removal from water by activated carbon may continue for
years. A fraction of NOM becomes irreversibly bound to GAC upon
adsorption(54). In a fixed-bed process treating water bearing both NOM
and more strongly adsorbed target contaminants, the NOM advances more
rapidly through the bed. This gives NOM the opportunity to foul or
preload the GAC, before the target contaminant is adsorbed. Preloading
effectively reduces the capacity and rate of target contaminant
adsorption(66). The longer GAC is preloaded by NOM, the less effective
will be the adsorption of target contaminants. This phenomenon may be an
important factor in determining the optimal design and operation of GAC
adsorbers, requiring that preloading duration be minimized in order to
achieve efficient target contaminant removal.
NOM preloading serves as a useful example of how adsorption models
are limited by their lack of testing, or verification, under appropriate
conditions. Because the time scale of NOM preloading is long, this impact
upon target contaminant adsorption is not observed in short laboratory
experiments typically used to calibrate adsorption models. Because model
testing (verification) has occurred under essentially the same conditions,
the need to account for preloading in order to provide accurate model
simulation has not been manifest. In fact, current adsorption models
appear unable to simulate the dynamics of preloading(42). The problem has
been revealed by the studies mentioned above, with models being tested
under a set of long-term field conditions allowing the effect of
preloading to become apparent. The question becomes, then, how to
incorporate this important effect into modeling predictions of target
contaminant adsorption.
Four strategies for modeling target contaminant adsorption under NOM
preloading conditions may be followed. First, the modeler may choose to
ignore the preloading effect a1together(51). Second, a lumped-parameter
approach may be taken, whereby the influence of preloading is accounted
for by modifying the adsorption parameters of the target contaminant(s).
Third, the lumped-parameter approach may be extended by treating the
target contaminant adsorption parameters as being functions of time or
throughput(42). This acknowledges that the impact of preloading is
expected to increase with the age of the adsorber. Fourth, the adsorption
model may incorporate the specific interactions between NOM, target
contaminant, and activated carbon. The adsorption models presently
available do not do so.
The constant, lumped-parameter approach was taken in this work. The
lumped-parameter approach provided the simplest means of modeling NOM
preloading, and of evaluating the impact of this factor upon DBCP
adsorption. This simple approach is also consistent with the present
limited understanding of the NOM preloading mechanism and its effect.

-------
6
MATERIALS AND METHODS
Adsorbate solutions were prepared in both organic-free water and in
ground water. Organic-free water (OFW) was obtained from a laboratory
water purification unit (Milli-Q). This high-purity water contained less
than 0.1 mg/1 of TOC, and had a resistance of greater than 18 Megohms-cm.
A ground water, collected from a well tapping the Great Miami aquifer of
southwest Ohio, was used as the natural water matrix for adsorption
experiments. TOC content of the ground water ranged from 1.9 to 2.5 mg/1,
with a pH of 7.6. The terminal trihalomethane (15 mg/1 chlorine added,
5-day reaction time at room temperature) was measured as 11 ug/L, and
terminal TOX (same reaction conditions) was measured as 80 ug/L. No
halogenated or aromatic VOCs were detected in the ground water. The
adsorbability of the organic matter in the ground water was determined by
bottle-point isotherm. The isotherm results indicated that about 0.9 mg/1
of the TOC (43% of the total organic carbon) was adsorbable. The maximum
amount of TOC adsorbed was about 10 mg TOC/gram activated carbon.
The DBCP used in all experiments was technical grade, 97% purity
(Pfatz & Bauer, Inc.). Stock solutions were prepared by transferring
neat DBCP by microliter syringe into 100 ml volumetric flasks containing
organic-free water, then agitating the flasks until the DBCP was
dissolved. Volumes of stock solution were transferred by pipet to achieve
desired DBCP concentrations in experimental solutions.
DBCP samples were analyzed by EPA Methods 502.1 (Purge and Trap/Gas
Chromatography with Hall ECD) and 504 (Microextraction/Gas Chroma-
tography). Analyses were performed by TSD personnel and by a contract
laboratory, Southwest Research Institute. Quality assurance data is
included with the experimental data in the Appendix.
The activated carbon used in this study was taken from a single
quantity of Filtrasorb F-400 (Calgon). The activated carbon was
pulverized for isotherm experiments and segregated into selected particle
size fractions for dynamic experiments. Powdered activated carbon (PAC)
was prepared by repeatedly crushing activated carbon until it passed a
100-mesh sieve (U.S. standard). The PAC was washed with organic-free
water in a beaker, and the non-settlable fraction decanted. Washing was
repeated until a clear supernatant was obtained. The PAC was then dried
at 105° C for 3 days. PAC was stored in Teflon-sealed bottles in a
desiccator. Granular activated carbon (GAC) was prepared by first
collecting activated carbon in the 20/40-mesh size fraction; a portion of
this was crushed, and the 60/80-mesh fraction collected from that. The
sizes of the two GAC fractions were each'Calculated as the average of the
passing and retaining sieve openings. This gave values for the GAC
particle radius of 0.0315 cm and 0.0107 cm for the 20/40 and 60/80 GAC,
respectively. Both GAC size fractions were washed to remove carbon fines;
drying and storage of GAC was the same as for PAC. The density of dried
GAC particles was determined, by displacement of water under vacuum, as
720 mg/cm.
A portion of the GAC was "preloaded" with natural organic natter by
extended contacting with ground water. The GAC was placed in a 10-cm
deep, 2.5-cm diameter column through which ground water was passed at a
rate of 100 ml/nin. This preloading was conducted for a duration of 10

-------
7
weeks. Preloaded PAC for isotherms was obtained by pulverizing wet
preloaded activated carbon and used directly. A sample of the preloaded
PAC was dried; it all passed a 100-mesh sieve. Preloaded GAC was
freeze-dried (Bionetics, Inc.), to allow handling, then sieved to collect
the 20/40 mesh fraction. A quantity of the 20/40 was crushed and
re-sieved to obtain a 60/80-mesh fraction, as well. Preloaded GAC was
stored in Teflon-sealed glass bottles in a desiccator.
Mathematical Adsorption Models
The pore-surface diffusion (PSD) model(22) was used to simulate DBCP
adsorption dynamics in batch and plug flow reactors. The choice of this
model for the present work was obvious, as the PSD is a readily available
and often-used dynamic adsorption model. The PSD model treats GAC
adsorption as a two-stage kinetic process, with mass transfer occurring
through the liquid boundary layer surrounding the adsorbent particle and
within the adsorbent particle itself. The PSD evolved from the HSD
(homogeneous surface diffusion) model, itself an outgrowth of the first
numerical model of adsorption dynamics to be applied in environmental
engineering, MADAM(4,5,6). Two parameters, kf, the film transfer
coefficient and Ds, the surface diffusion coefficient, quantify the two
steps of mass transfer resistance incorporated in the HSD model. The PSD
model is similar to the HSD, except that in addition to surface diffusion,
intraparticle mass transfer includes the parallel process of pore
diffusion. The PSD model, then, makes use of three kinetic parameters:
kf, Ds, and Dp, the pore diffusion coefficient. Both the HSD and PSD
models allow for adsorption equilibria to be described by any desired
isotherm equation, including multicomponent isotherms(9). For
single-component simulation of DBCP adsorption, equilibrium was described
by the Freundlich isotherm equation.
The batch reactor adsorption model was developed and implemented on a
Hewlett Packard 9000-series microcomputer. An ordinary differential
equation solver, based upon an algorithm of Michelsen(37), was also
developed to run the batch models. The plug flow reactor adsorption model
was obtained from Michigan Technological University and run on a VAX
11/785 minicomputer.
Experimental Apparatus
A number of experiments were conducted to assess the equilibrium and
kinetics of DBCP adsorption, for a variety of experimental conditions.
The PSD model was used as a vehicle to analyze and interpret the dynamic
adsorption data. The resulting adsorption parameters were used to
calibrate the model, so that performance predictions of DBCP adsorption
could be made.
Adsorption Isotherms
The purpose of isotherm experiments is to determine the equilibrium
achieved between adsorbent-phase and aqueous-phase adsorbate
concentrations(45). Isotherm experiments were conducted in closed 1-L
bottles, wherein PAC is contacted with adsorbate solution. Both the
adsorbate concentration and the mass of PAC were varied between bottles,

-------
8
so that adsorption equilibrium over a range of concentrations could be
evaluated. Adsorbent-phase concentration was calculated by mass balance.
DBCP adsorption isotherms were developed for three conditions: virgin PAC
in organic-free water, virgin PAC in ground water, and preloaded PAC in
ground water. Organic-free water isotherms were conducted by DWRD(41).
The procedure followed for DBCP isotherms in organic-free and ground
water solutions was as follows. Preweighed doses of PAC were added to the
isotherm bottles, which were then filled with adsorbate solution. The
bottles were sealed, then agitated on a rotating apparatus until
equilibrium was achieved. Agitation was stopped, the PAC was allowed to
settle from solution, and samples of equilibrated adsorbate solution were
collected for analysis.
For the isotherm experiments conducted with preloaded PAC the
procedure was modified. Preloaded, wet activated carbon was pulverized
and added to isotherm bottles directly. Dosage was determined by
collecting the PAC on filter paper at the conclusion of the experiment.
Control tests indicated that 99% of PAC added to isotherm bottles was
recovered by filtration.
To determine the length of time necessary for isotherms to reach
equilibrium, a time-to-equilibrium experiment was conducted. Results
(Figure 1) indicated that DBCP adsorption equilibrium was achieved in
about 7 days in ground water solution. The equilibration time for
organic-free and ground water isotherms was 13 days, and for the preloaded
isotherm, 27 days. Losses of DBCP from isotherm bottles containing no
activated carbon were negligible.
Kinetic Experiments
Differential-column batch reactor (DCBR) and short-bed adsorber
experiments were conducted to evaluate the kinetics of DBCP adsorption. A
dynamic verification column (DVC) experiment was also performed, to
provide data against which to judge the accuracy of predictions made by
the adsorption model.
DCBR - The DCBR has been used extensively to study adsorption
kinetics of VOCs as well as humic materials(26). The purpose of DCBR
experiments is to determine intraparticle mass transfer rates. Adsorbate
solution is withdrawn from a completely-mixed vessel, and a portion is
passed through a thin bed of GAC and then returned to the vessel. Flow
through the bed is maximized in order to (eliminate mass transfer^
resistance external to the GAC particles. The experiment is monitored by
collecting samples from the vessel over the duration of the experiment, 1
to 2 weeks. The same three adsorbate/adsorbent conditions
(DBCP/OFW/Virgin AC, DBCP/GW/Virgin AC, and DBCP/GW/Preloaded AC) for
which isotherms were developed were run in DCBR experiments. The design
and conduct of DCBR experiments followed the description of
Dobrzelewski(15), with some modifications. Principally, the volume of the
batch reactor vessel was increased to 47 liters, due to the high affinity
of DBCP for activated carbon. Tracer studies ensured that
completely-mixed conditions were maintained in the DCBR vessel.

-------
9
SBA - The SBA has been used principally to study the adsorption
kinetics of phenolic compounds and pesticides(32,51). Both intraparticle
and external (film) mass transfer rates may be determined from SBA
experiments. The SBA is a column breakthrough experiment in which, as its
name implies, a shallow depth of GAC is used. The depth of GAC must be
sufficiently short to allow immediate adsorbate breakthrough in the column
effluent. SBA breakthrough is monitored until the column is exhausted.
Two classes of SBA experiments were carried out. In the first, two SBA's
were set up to duplicate the hydraulic conditions (flow rate and GAC size)
of the differential column experiments. In the second, SBA hydraulic
conditions duplicated that of the DVC.
DVC - A longer column experiment (in terms of both column depth and
run duration) was conducted concurrent with the SBA. This experiment
produced a DBCP breakthrough profile which was used to verify the
predictive ability of the PSD model. DBCP-ground water solution was fed
from a covered 100-gallon stainless steel tank via a gear pump to the SBA
and DVC, arranged in series. The solution passed through a sand bed
before entering the columns. Flow was controlled by a needle valve and
measured by collecting column effluent in a graduated cylinder for 1
minute. Samples were drawn from valves located before and after the SBA
(influent and SBA effluent) and after the DVC (DVC effluent).
RESULTS AND DISCUSSION
Experimental conditions and data are presented, for each experiment,
in the Appendix.
Adsorption Isotherms
Results of the adsorption isotherm experiments are presented in
Figure 2. The data were fitted to the Freundlich isotherm equation:
Qe = Kf* ce "
where: Ce = equilibrium liquid-phase adsorbate concentration
Qe = equilibrium solid-phase adsorbate concentration
Kf = Freundlich constant
n = Freundlich exponent
which is linearized on a log-log plot:
i
log (Qe) = log (Kf) ~ n*log (Ce)
The Freundlich equation was fitted to the three sets of isotherm
data, by linear least-squares regression of the log-log data. The
resulting isotherm parameters are presented in Table 2. ^The two
highest-concentration data points from the preloaded/GW isotherm were
excluded from regression, as they exhibited a pronounced deviation from
linearity. These data points correspond to isotherm bottles containing
the smallest amounts of activated carbon. Multicomponent equilibrium
theory, such as IAST(9), suggests that competing adsorbates may cause the
capacity of the target compound to decline in isotherms conducted with low

-------
10
activated carbon dosages(43). Some fraction of the NOM in the ground
water may be expected to compete with DBCP for adsorption. Therefore,
these two data points were treated as outliers from the isotherm.
The slopes of the Freundlich isotherms are nearly parallel, but
capacities (amount adsorbed at a given liquid-phase concentration) varied
by as much as an order of magnitude. Some reduction of capacity was
observed when DBCP was adsorbed from ground water instead of organic-free
water. However, the drop in capacity was much more evident for the
isotherm conducted with preloaded activated carbon. This parallel,
downward shift in adsorption isotherms has also been observed for
trichloroethylene adsorbed onto activated carbon preloaded with Rhine
River water(66). In that study, the downward shift in the isotherm
increased after longer preloading durations.
Several methods may be used to describe the ability of isotherm
equations, such as the Freundlich, to fit the data. The regression
correlation coefficient and 95% confidence intervals are commonly used.
However, it is also of interest to examine the sources of error in the
isotherm data itself. A nonparametric approach of examining the bounds of
isotherm data errors was taken here. For a first approximation, the
source of error was assumed to arise solely from chemical analyses, with
analytic error limited to 10%. The 10% error was added to and subtracted
from the initial and residual DBCP concentration values from the isotherm
data, and the amount adsorbed was calculated by mass balance:
0e = (V/M)*( C0-Ce )
where: V/M = inverse of activated carbon dosage
C0 = initial liquid-phase adsorbate concentration
Error in the amount adsorbed is contributed by errors in both CD
and Ce. When Ce approaches the value of C0, the limits of error in
the amount adsorbed will grow quite large. The results of the error
analysis are displayed in Figure 3 where the 10% error bounds are
indicated by bars around the isotherm data. Clearly, uncertainty in
adsorption capacity may grow quite large at the high-concentration end of
the isotherm and, in fact, this is often observed. The 95% confidence
bounds on the isotherm regressions are also plotted in the figure. It
should be noted that the confidence bounds, based upon the linear
regression of isotherm data, give no indication of this behavior. The
error-bound evaluation of isotherms may be useful for checking data for
outliers. It may also assist in the design of isotherm experiments to
avoid the undesirable spread in isotherm data due to error, by determining
appropriate ranges of initial and residual adsorbate concentration.
DCBR
Six successful DCBR runs were nade, with experiments falling in two
groups. In the first group, DCBR1, 2, 3 and 4, the design and conduct of
the experiment was optimized, and the impact of concentration and particle
size upon intraparticle mass transfer was checked. In the second group of
experiments, DCBR4, 5 and 6, the effects of the ground water matrix and
preloading upon DBCP adsorption kinetics were evaluated.

-------
11
Analysis of the DCBR data was accomplished by fitting PSD model batch
concentration profiles to the experimental data, until the best agreement
between data and aodel prediction was obtained. The fitting procedure was
automated via computer program, which varied one or more kinetic^
parameters while searching for the minimum of an objective function. The
objective function, a relative sum-of-square error (Rel.SSE), was defined
as:
D [ (Cd • t « , 1~Cao d•1.I)/Cd •t•, 1 ]2
Fel.SSE = I 	
i=l	n-1
The relative SSE normalizes each squared error by the experimental
value, reflecting the assumption that the error in concentration
Measurements is proportional to the concentration(13).
The first four runs were used to optimize the DCBR experiment, as
well as to check the effect of varying DBCP concentration and GAC particle
size upon intraparticle mass transfer. The DCBR experiment was modified
for DBCP, which has different equilibrium and kinetic adsorption
properties than compounds previously investigated by this technique.^ In
comparison to the adsorbates studied by Dobrzelewski(15), DBCP exhibited a
greater adsorption capacity and slower kinetics. These adsorption
properties necessitated using a larger ratio of adsorbate volume/adsorbent
mass, in order to prevent excessive depletion of DBCP in batch solution.
A smaller GAC particle size was chosen to compensate for the slower
kinetics of DBCP. Using 60/80-mesh GAC instead of the commercial
12/40-mesh size was expected to reduce the duration of the DCBR run by a
factor of 20, assuming intraparticle mass transfer limitation.
Unfortunately, these modifications made the DCBR results difficult to
analyze. The necessary flow velocity to overcome film transfer resistance
could not be attained in DCBRs with 60/80-mesh GAC. As a consequence, the
data could not be analyzed for intraparticle mass transfer rate alone.
Instead, rates of both film transfer and intraparticle mass transfer had
to be searched from the data. This leads to the possibility of covariance
between the two parameters diminishing the accuracy of the DCBR results.
The treatment of the 60/80-mesh data, to overcome this drawback, is
considered below. The experimental problem was resolved by switching to a
larger GAC particle size, 20/40-mesh, which allowed film transfer
resistance to be minimized in the DCBR. The second principal change made
in DCBR experiments was to move the sampling location. Initially, samples
of batch concentration were taken from the recirculating flow. Sampling
in this manner disrupted the flow, and was awkward to perform. In latter
DCBR experiments, samples were collected from the batch vessel with a
50-ml syringe and Teflon spaghetti tubing.
The experimental conditions and results for the DCBR experiments are
presented in Tables 3 and 4, respectively. Table 5 presents the
Intraparticle mass transfer parameters (MTPs) determined from each
experiment, along with the associated 95% confidence interval. DCBR data
and best-fit model curves are plotted in Figures 4-9. The confidence
interval is a measure of precision attained in fitting model predictions
to the data(16). The confidence intervals were obtained from plots of Ds

-------
12
versus Re1.SSE(26). The plots are presented in Figures 10-15. Table 4
also presents values of the dimensionless Biot Number (Bi) and Stanton
Number (St). These parameters indicate whether the DCBR has been properly
designed to evaluate adsorption mass transfer. Bi is the ratio of
Intraparticle to film transfer resistance; a value of 30 or greater
indicates intraparticle control of mass transfer, the desired condition
for the DCBR(15). St relates the length of the mass transfer zone to the
adsorbent depth; a value less than 0.017 has been suggested for proper
DCBR operation(15). The 20/40-mesh DCBR experiments were better able to
¦eet these two conditions than were the 60/80-mesh experiments. Mapping
Rel.SSE as contour surfaces, Figures 16-17, confirms that model analysis
of the 20/40-mesh DCBR data is sensitive to Ds and insensitive to kf.
Both the results of single-parameter (Ds) and 2-paraneter (Ds and kf) DCBR
analysis are presented for DCBR1-4. As will be discussed below, the
single-parameter analysis was favored over the 2-parameter as a
data-fitting procedure. Figure 18 provides a visual comparison of
best-fit Ds values and their 95% confidence intervals, for all DCBR
experiments.
In DCBR1-4, the impact of varying GAC particle size and initial DBCP
concentration upon Ds was checked, because these variables have been
reported to cause variation in surface diffusivity(49,53). Such variation
is significant, as non-constant MTPs make extrapolating model predictions
from one set of conditions to another difficult. The impact of varying
DBCP concentration was checked in DCBR3 (11 ug/L). A one-tailed T-test
indicated that DCBR3 resulted in a significantly lower Ds than did DCBR1
and 2 (Co=80 ug/L). The critical region was T>6.31; the value of the test
statistic, t, was 11.4. The impact of varying GAC particle size was
checked in DCBR4 (20/40-mesh). The value of Ds from DCBR4 was smaller
than the values from DCBR1 and 2 (60/80-mesh), although the difference
could not be statistically tested, as the Ds sample values did not share a
common variance. The standard deviation of Ds values for 60/80-mesh DCBR
experiments was significantly greater than for 20/40-mesh experiments, due
to the improved "resolution" of parameter fitting to data obtained from
20/40-mesh DCBR experiments. DCBR1 and DCBR2 were replicate DCBR
experiments, performed to check the reproducibility of the experiment and
its results. The coincidence of Ds values indicates a good agreement
between the duplicate DCBR experiments.
The results of DCBR4 and 5 show there to be no change in Ds when DBCP
is adsorbed from ground water as opposed to OFW. Analyzing the data from
DCBR5 (GW-virgin GAC) and DCBR6 (GW-Preloaded GAC) for intraparticle mass
transfer indicates only a slight decline in Ds. The Ds value fitted to
DCBR6 data is not significantly different than the Ds value for DCBR5 or,
for that matter, DCBR4. The change in the values of the Biot number is
due to declines in DBCP adsorption capacity for GW and Preloaded/GW
conditions.
It was found that fitting DCBR6 data by modeling pore instead of
surface diffusivity resulted in an equally good description of the data.
Intraparticle transfer rate was reduced to the point where pore diffusion
could compete with surface diffusion as the controlling mechanism of mass
transfer resistance. In fact, from the results of DCBR6, it cannot be
determined whether intraparticle mass transfer is taking place by surface

-------
13
diffusion, pore diffusion, or some combination of the two. Surface
diffusion cannot be presumed as the predominant form of intraparticle mass
transfer in DCBR6. In contrast, surface diffusion was clearly predominant
in DCBR1-5, based upon model sensitivity analysis.
To analyze the 60/80-mesh DCBR data, kf as well as Ds was fitted to
the data by the HSD model. Poor results were obtained from the
two-parameter data analysis, as indicated by wide 95% confidence
intervals for Ds and by large variation between kf values for runs 1, 2
and 3. Because the flow conditions were identical, equal kf values were
expected. Subsequent to this finding, PSD model sensitivity analysis was
performed, indicating that kf values could not be adequately determined
from the DCBR data. To accurately search Ds from the 60/80-mesh DCBR
data, an independently determined value of kf was needed. This value was
determined from an SBA experiment, described below. Kf was then fixed as
the experimental value, and the DCBR data was searched for Ds alone. The
single-parameter search proved to be an acceptable means of analyzing the
DCBR data for Ds. Pore diffusion was found to have no impact upon DBCP
adsorption kinetics in organic-free water.
In DCBR4, 5 and 6 the effect of water matrix and activated carbon
preloading upon intraparticle mass transfer was evaluated. Film transfer
resistance was minimal in DCBR4, and eliminated in DCBR5 and 6, so that it
was necessary to search only the intraparticle mass transfer parameter
from these data.
The accuracy of MTPs determined by fitting the adsorption model to
DCBR data depends, in turn, on the accuracy of the other parameters used
as input to the model. This source of imprecision is not reflected in the
95% CIs presented above. In particular, GAC particle size and adsorption
isotherm parameters are critical to the fitting of DCBR data. GAC
particle size is conventionally found by sieve analysis, with an average
value of particle diameter determined from the results. This value is
then input to the model, where 1t is treated as the dimension of a
spherical adsorbent particle. Several researchers have questioned the
suitability of this approach, citing the nonuniform shape(46) and
polydisperse distribution of particle sizes(60) as confounding influences
upon film transfer. To check the shape and size of GAC particles used in
this study, samples of 20/40- and 60/80-mesh GAC were examined beneath a
microscope. The GAC was found to be roughly cylindrical in shape. For
both 20/40 and 60/80, the median dimension of 50 particles sized by
microscope was greater than that determined via sieving (the average of
sieve size openings for passing and retaining sieves) and, in fact, was
slightly greater than the size of the opening in the passing sieves. The
difference between average sieve size and median microscope size may be
explained by the roughly cylindrical shape of the GAC, allowing it to pass
more readily through sieves, and by the orientation of GAC on the
microscope grid, which may tend to show the longer dimensions of the
three-dimensional particle. Which value for particle size 1s more
accurate for modeling purposes was not determined. If consistent size is
used for both model calibration and application, errors arising from
inaccurate particle diameter will be compensated by the MTP's.

-------
14
The impact of particle size determined by the two methods upon the
value of Ds searched from the DCBR data was evaluated. Optimization of
DCBR 5 data was performed with r values determined by sieve analysis
(r=0.0315 cm) and by microscopic sizing (r=0.D446 cm). The resulting
best-fit values of Ds were 4.5E-11 and 8.8E-11 cm*cm/sec, respectively.
Obviously, the determination of Ds from DCBR data depends upon the value
of r used, which in turn depends upon how r is measured.
The parameter-fitting is similarly dependent upon the values chosen
for isotherm parameters. A sensitivity analysis of Ds-fitting, similar to
the one used for r, was applied to the Freundlich isotherm constant,
K<. For DCBR 5 data, a 25% decrease in Kf resulted in a doubling of
tne optimum Ds value. The accuracy of isotherm parameters in describing
the equilibrium achieved in the DCBR experiments was checked by mass
balance. For all DCBRs, reasonable agreement was achieved between DCBR
equilibrium and isotherm predictions. This agreement is demonstrated in
Figure 19, by comparing DCBR equilibrium to the adsorption isotherms. For
several of the DCBR experiments, equilibrium was not attained, and the
"equilibrium" points plotted in Figure 19 were calculated based upon the
final concentration. This results in an underestimation of equilibrium
capacity.
Systematic error in analyses was a confounding influence on MTP
fitting of DCBR data. Day-to-day variability in GC performance introduced
correlated errors to the DCBR data, and possibly biased the data
regression. Figure 20 shows DCBR6 data, before and after the analytic
results were "corrected" by the QC recoveries for each day's analyses.
The data, as received, apparently contains a non-random error component
due to analytics. Because GC performance may vary within one day as well
as between days, the "correction" of data in this manner was not
considered valid. In all cases, data was analyzed "as received."
Apparently, tighter analytic quality control is appropriate for
experiments in which time-series data is analyzed.
Short-Bed Adsorber
Two groups of SBA experiments were carried out. The first, termed
high-flow SBAs, were conducted at packed-bed flow conditions duplicating
the differential column of the DCBR experiments. The high-flow SBAs
provided independent film transfer coefficient values for use in DCBR data
analysis, as was mentioned above. The second, identified as a long-term
SBA, was run to provide MTP estimates for the model verification aspect of
the project.	(
The high-flow SBAs were designed to duplicate the GAC contacting
which took place in the DCBR, thereby allowing an independent estimate of
the film transfer coefficient to be made. The experimental conditions of
the high-flow SBAs are listed in Table 5. The 60/80-mesh high-flow SBA
conditions correspond to the differential column hydrodynamics of DCBR 1,
2, and 3. The 20/40-mesh high-flow SBA duplicated the hydrodynamics of
DCBR 4. Only the initial portions of SBA breakthrough were produced in
the high-flow experiments, because only kf was of interest. The results
of fitting the HSD model to the high-flow SBA data, giving an optimum
value of kf, are also presented 1n the Table. The specification of a

-------
15
value of intraparticle Bass transfer (Ds or Dp) had no impact upon the kf
value obtained. Figures 21-22 compare the high-flow SBA data to best-fit
model simulations.
In order to use the results of the high-flow SBA to provide a kf
value for DCBR 5 and 6, where the differential column flow was 2.5 times
greater, the value was scaled by using a correlation of Gnielinski*s(24).
The correlation relates film transfer to packed-bed hydraulics.
Gnielinski's correlation indicates that film transfer increases 1.5 tines
due to the increase in flow. Two other film transfer correlations, of
Williamson(65) and 0hashi(40), estimate the increase as 2.0 and 1.4,
respectively. Because film transfer resistance was eliminated in DCBR 5
and 6, such an estimate of kf was more than sufficient.
The long-term SBA was designed to give a good estimate of
intraparticle mass transfer resistance for the DBCP/GW/Preloaded GAC
system under column operation. Preliminary analysis of DBCP kinetics
Indicated that intraparticle mass transfer would be controlling at
full-scale conditions. Because the long-term SBA was run until
exhaustion, the breakthrough data could be searched for parameter values
for both film transfer and intraparticle mass transfer(32). The
two-parameter search used the same objective function, the relative SSE,
that was used in fitting the DCBR data. Experimental conditions and
results are presented in Table 6. The long-term SBA breakthrough data and
best-fit model simulations are plotted in Figure 23.
As was the case for DCBR 6, the HTP fitting of the long-term SBA
could be accomplished by varying either Ds or Dp as the coefficient of
intraparticle mass transfer. Best-fit simulations were virtually
identical. The values of Ds and Dp from SBA analysis were nearly
identical to the values from DCBR 6, indicating that the two methods yield
equivalent results for intraparticle resistance. Long-term SBA sampling
under-represented the initial stage of breakthrough, due to an error made
in sample collection. This resulted in wide confidence intervals about
the mass transfer parameters. Figures 24 and 25 present error contour
plots for the fitting of two parameters.
Two experiments used to determine intraparticle mass transfer
resistance, the DCBR and SBA, gave virtually identical results for the
kinetics of DBCP adsorption on preloaded GAC. Each experiment has
advantages making it more attractive for a particular application. The
DCBR uses a much smaller volume of adsorbate solution, which may be
important when adsorption tests are performed with a sample of the actual
water to be treated. The SBA may be completed more rapidly, and allows
the simultaneous determination of external (film) as well as intraparticle
mass transfer rates. Errors in the SBA mass transfer parameters will be
compensating(32), making model predictions more accurate than would be
obtained in predictions made with independently-determined mass transfer
parameters. This advantage is lost, however, if model simulations are
made at hydrodynamic conditions different from the SBA.

-------
16
Dynamic Verification Column
The DVC was run concurrently with the SBA to produce a breakthrough
profile similar to that of a full-scale adsorber. Experimental conditions
for the DVC were identical to the long-term SBA, except that the DVC was
considerably longer. The DVC experimental conditions in Table 7 and the
influent and effluent data shown in Figure 26. Some variation in influent
DBCP concentration was observed, producing an irregular breakthrough
profile.
The DVC data were used to test the predictive ability of the PSDM to
model DBCP breakthrough. Specifically, the DVC tested model prediction of
the adsorption of DBCP from ground water solution onto preloaded GAC.
This condition was of greatest interest, because it was selected to
duplicate actual water treatment conditions. The DVC data are compared to
the model prediction of DVC breakthrough in Figure 26. For the
verification model prediction, DBCP/GW/Preloaded GAC isotherm parameters
were used. MTPs were determined in the long-term SBA experiment. As was
the case for long-term SBA calibration, model predictions with surface and
pore diffusion intraparticle resistance give nearly equivalent results for
the DVC prediction. Agreement between the DVC data and PSD model
prediction is fairly good, with error statistics (Rel.SSE) of 0.1794 and
0.0927 for pore diffusion and surface diffusion predictions,
respectively. Some systematic deviation between the data and model
prediction occurs between 2500 and 9000 minutes in the DVC run, where the
model overpredicted the breakthrough concentration of DBCP. The DVC
sampling may not have adequately tracked the influent DBCP concentration,
possibly influencing the model prediction error.
This verification step is used to determine how well the model
prediction fits independent data and, hence, the degree of confidence that
should be placed with model predictions. If the model prediction does not
closely match the verification data, then other model predictions are
likely to be inaccurate as well. Although perfect agreement between model
prediction and DVC data was not observed, the deviation was such that
model predictions were conservative (rapid breakthrough), so model
verification was considered successful. Because the surface diffusion
model prediction was more accurate than the pore diffusion model
prediction, surface diffusion model predictions will be used for further
model applications in this study.
Sensitivity analysis was performed upon the DVC model simulations.
Three PSD model parameters, K<, Ds, and kf were varied by 50% and the
resulting model simulations of DVC breakthrough compared to the
simulation made with the experimentally-determined parameters. The
results are plotted in Figures 27-29. This analysis indicated that the
model was most sensitive to equilibrium, and not kinetic, factors. DVC
predictions were observed to be most sensitive to the Freundlich isotherm
constant, Kf. Surface diffusivity had some effect upon the model, but
sensitivity to film transfer was negligible. The model was also sensitive
to the influent profile. By reducing the influent concentration values by
10% during the first phase of the DVC simulation, model fit to the data
was considerably improved. This is shown in Figure 30. It is possible
that a more accurate description of the DVC influent may have allowed for
a better model simulation to be made.

-------
17
The verification of the PSD model against DVC data indicates the
importance of Incorporating the effect of preloading in the model
prediction. This point may be demonstrated by comparing the DVC data to
three different PSD model predictions. The first is the simulation based
upon adsorption parameters determined for the DBCP/GW/Preloaded GAC
system. The second is based upon parameters for the DBCP/GW system, and
the third, for the DBCP/OFW system. The comparison is shown in Figure
31. If preloading effect is not included in the model simulation,
breakthrough will occur significantly earlier than predicted. Such
nonconservative prediction errors are highly unsatisfactory. The simple
lumped-parameter modeling approach allowed for reasonably accurate model
simulation under NOM-preloading conditions.
The modeling of DBCP adsorption from ground water onto preloaded GAC
was of primary interest in this study, because this condition was chosen
to simulate actual water treatment conditions. As indicated by the
successful calibration and verification described above, DBCP adsorption
under simulated water treatment conditions can be modeled by a simplified
method. In this method, the single-component PSD model is used to predict
breakthrough of DBCP, the target contaminant. The parameter values used
in modeling are determined by rapid, small-scale experiments employing GAC
which has been preloaded with natural organic matter. The DBCP adsorption
capacity of preloaded GAC was observed to be an order of magnitude less
than the capacity of virgin GAC. Preloading was not found to
significantly affect GAC adsorption kinetics. The simplicity of this
modeling approach is appealing, as it allows the impact of NOM preloading
upon target contaminant adsorption to be readily determined.
The limitations of this modeling method should be considered as
well. First, verification of modeling predictions at full-scale, water
treatment conditions has not been performed. Verification was performed
with a data set assumed to approach those conditions. The dynamic
verification column experiment was an attempt to model the full-scale
adsorber, within the constraints of a short, laboratory experiment.
Obviously, a full-scale adsorber data set would have provided a much
better verification data set. Procedural aspects of the DVC experiment
may be questioned as well. Particularly, the handling of preloaded GAC
was not fully evaluated: freeze-drying and re-sieving could have altered
adsorption behavior. Handling in this manner should not affect the NOM or
activated carbon. However, a change in the adsorption bond or
configuration of NOM to the activated carbon as a result of freeze drying
is possible. Agreement of the preloaded isotherm to the capacity observed
in kinetic experiments indicates that at least NOM impact upon DBCP
adsorption equilibrium was not seriously affected. Second, the impact of
NOM preloading upon adsorption capacity and rate was determined at a
constant (10-week) point. The impact of preloading upon target
contaminant adsorption is expected to be a function of the extent of NOM
"loading" which occurs prior to the target contaminant reaching a given
depth in the GAC bed. Thus, preloading will not have a constant impact
(as is assumed by this model) but will rather vary, increasing with
contacting time and depth in the bed(42). Whether this, or a more
complicated model, is appropriate for modeling NOM preloading impact upon
target contaminant adsorption will depend upon the constancy or variation
of the preloading effect. The data available at present are not
sufficient to make such determination.

-------
18
Adsorption OTP Correlations
Correlations may be found In the literature for estimating mass
transfer paraneters, including kf, Ds, and Dp. Correlations for the film
transfer coefficient were used in the analysis of the high-flow SBA data
(above). In general, correlations provide a simple, expedient alternative
to kinetic experiments in determining MTPs for adsorption modeling.
However, the accuracy of correlations is uncertain. The agreement between
correlation-based HTP estimates and values determined experimentally is of
considerable interest, and was examined as a part of this study.
There are a significant number of correlations for the film transfer
coefficient in the chemical engineering literature. Three correlations
for kf which have been applied to GAC are Williamson's(65),
Gnie1inski's(24), and Ohashi's(40). All are functions of the form(46):
kf = f (Di	,SC)
where: Di = free liquid diffusivity
t = void fraction of adsorbent bed
R. is the Reynolds Number and Sc Is the Schmidt Number, as
defined by Roberts, et al-(46) These dimensionless parameters quantify
the flow conditions of the packed bed and the diffusivity of the
adsorbate, respectively. These correlations were based upon data for
spherical particles. Several 1nvestigators(46,60) have proposed modifying
the correlations to account for the irregularity and roughness of GAC
particles, which enhance the rate of film transfer. Williamson's
correlation has been found to provide accurate GAC kf values for typical
adsorber flow rates(lO) (Re from 1 to 10). Gneilinski's correlation,
when corrected for particle sphericity, reportedly provides the most
accurate relationship for kf over a wide range of flows(46) (Re from 1
to 10,000).
The kf values calculated from the three correlations are presented
for comparison to experimental DBCP values in Table 8. At the high-flow
conditions, there 1s reasonable agreement for the 60/80-mesh particle
size. The correlated values of kf for 20/40-mesh GAC are more scattered,
with Williamson's coming closest to the experimental result. The
correlations provide better kf estimates at the low-flow condition of the
long-term SBA, which Is within the range jof flow (Re of 1-10) most
typical for adsorber operation.	|
Correlations for Ds have been proposed by Suzuki, et al.(55), and by
Dobrzelewski(15). The Dobrzelewski correlation is based upon adsorption
data for a number of chlorinated alkanes, alkenes, and aromatic compounds,
and will be considered here. The form of the correlation Is:

-------
19
Ds e PDFC * PSDFR
where PDFC is the pore diffusion flux contribution, as defined by
Crittenden(7) and PSDFR, the pore-to-surface diffusion flux ratio, has
been empirically determined as approximately 6.58. A notable feature of
this correlation is that it treats Ds as being dependent upon the
equilibrium distribution between aqueous and adsorbent-phase
concentrations. For favorably adsorbed compounds such as DBCP, the
correlation indicates a decline in Ds as concentration drops. The
possibility of a concentration-dependent Ds has been suggested(53).
Ds values calculated from the Dobrzelewski correlation are presented
for comparison to experimental DBCP values in Table 9. Agreement between
the correlation and experimental values is also plotted in Figure 32. In
the case of OFW experiments (DCBR 1-4), the correlation values fell within
the 95% confidence interval for the experimentally-determined Ds values.
Apparently, the Dobrzelewski correlation is valid for DBCP adsorption from
organic-free solution. For the GW solution and GW/Preloaded GAC
experiments (DCBR 5 and 6; Long-term SBA), a systematic deviation of
experimental values from the correlation was observed. The experimental
Ds values were smaller than the correlation predicted. Agreement with the
experimental Ds values for DBCP in GN matrix was obtained by changing the
value of the PSDFR from 6.58 to approximately 1.71.
The pore diffusion coefficient, Dp, is usually calculated as(22):
Dp s Di / Tp
where tp is the tortuosity of the pores within the GAC. By setting
intrapafticle tortuosity equal to 1, the maximum theoretical value of Dp
is.obtained. For DBCP, the maximum Dp value is therefore 8.2E-06
cnr/second at 25° C. This value is indeed greater than the
experimental Dp value of 3.7E-06 (DCBR 6) and 4.0E-06 (long-term SBA).
The back-calculated value of intraparticle tortuosity, by the relationship
above, is 2.1 to 2.2. Values of 0_, reported in the literature, range
from 2 to 6 for target contaminant adsorption onto activated carbon(lO).
The agreement between experimental and reported values of tp confirm the
applicability of the Dp correlation.
At this point, most studies of adsorption modeling in the literature
end(4, 5, 6, 7, 27, 32, 36, 44, 51, 52, 54, 57, 60, 62, 66), the model and
parameter estimation methods having been verified, with information
regarding the adsorption process inferred from the exercise. While it is
important to advance the understanding of adsorption as a process, this
misses the point that the resulting model is a tool of some practical use.
In this light, calibration and verification is only a prelude to putting
the model to work.
MODELING APPLICATIONS
The most important question to the engineer or regulator, with
regards to adsorption modeling, Is what can it be used for? To address
this question, model applications will be demonstrated by example.
Namely, the adsorption model will be used to examine the impact of several
design alternatives upon performance. By coupling the adsorption model to

-------
20
a model of treatment cost, the cost of adsorption will be examined as
well. Simulations of DBCP adsorption will be made with the PSD model,
calibrated to isotherm and kinetic data developed in this study, and
scaled up to full scale conditions.
The treatment objectives for DBCP adsorption were selected as
follows:
Influent DBCP Concentrations - 2, 5 and 20 ug/L
Effluent DBCP Concentrations - 0.1, 0.2 and 1 ug/L
DBCP occurrence data suggests that the lower influent concentrations
are sore likely levels of contamination. Because the proposed MCL for
DBCP is 0.2 ug/L, that concentration is the treatment objective of
greatest interest. The calculations of treatment cost were based upon a
system sized to provide 3.25 MGD average flow and 7.52 MGD maximum
(design) flow. Treatment costs for other system sizes may be estimated
from the same set of adsorption model simulations utilized here.
The PSD model predictions were based upon adsorption parameters
developed for the DBCP/GW/Preloaded GAC system, as described in previous
sections of this report. Mass transfer parameters were scaled to water
treatment conditions. The film transfer coefficient, kf, was calculated
from Williamson's correlation. The surface diffusion coefficient, Ds, was
calculated from Dobrzelewski's correlation, modified to fit the
experimentally-determined Ds values for DBCP. Thus, Ds was dependent upon
the influent DBCP concentration. Any error in correlated Ds values is
expected to produce conservative model results(10). Isotherm parameters
were not modified for full-scale simulations. Physical properties of the
GAC were those of commercial-grade (12/40-mesh) Filtrasorb 400(15).
Full-scale adsorbers were configured as 12-ft diameter, vertical,
cylindric, steel pressure contactors. Three adsorber empty-bed contact
times (EBCT; 5, 7.5 and 10 min.) and two adsorber stagings (single and
two-in-series) were considered, resulting in a total of six possible
adsorber "train" configurations. To meet design flow, nine adsorber
trains would be in operation. Only four adsorber trains would be in
operation at average flow.
DBCP adsorption was simulated for three influent concentrations,
three treatment objectives, and six train configurations. All necessary
computer runs were completed in a 1-hour timesharing session on the VAX
computer. Model parameters used in the simulations are listed in Table
10. Adsorption mass transfer was found to be controlled by surface
diffusion, resulting in fairly broad breakthrough profiles, one of which
is displayed in Figure 33. Because film transfer did not influence
adsorption kinetics, varying flow velocity (and hence, kf) had an
insignificant effect upon performance. Performance predictions were
converted to carbon usage rates (UR; kg GAC used/million liters treated)
for comparison and use in cost estimates. For single-stage adsorbers,
usage rate was calculated as:

-------
21
UR = W /(t*Q)
where
W = Weight of 6AC in Adsorber
t = Tine in Service Until Effluent
Meets Treatment Objective
Q = Flow Rate Through Adsorber
For two adsorbers operated In series, UR was calculated in a
different manner. In this configuration, countercurrent movement of
adsorbate and GAC is approached by removing the first adsorber in series,
directing influent flow to the second (which now becomes the first), and
adding a fresh adsorber to the second position. This cycling of adsorbers
is performed when the second column effluent concentration reaches the
treatment objective. The series configuration allows for fuller
utilization of GAC, potentially reducing the cost of treatment(31,38). In
order to estimate UR for series operation it was assumed that, following
an initial start-up phase, adsorption would approach constant-pattern
conditions(8). At constant-pattern conditions, the degree of GAC
exhaustion in an adsorber may be readily determined from the effluent
breakthrough(27). The conditions being modeled here should not violate
any of the conditions necessary for constant-pattern adsorption; any
alternative means of forecasting series adsorber performance would be
significantly more complicated. For adsorbers in series, then, UR was
calculated as:
UR = fs * URe
where	fs = Fractional Capacity Attained in
First Series Contactor
URe = Cj/qe
(Usage Rate Based Upon Isotherm)
Ci = Influent Adsorbate Concentration
qe = Kf * q n
(fclAC Capacity 1n Equilibrium with C^)
The fs is based upon the constant-pattern solution to the HSO
model. The throughput (dimensionless time or volume treated) exiting the
first adsorber at the moment it is removed from service (triggered by the
second adsorber effluent reaching the treatment objective) was calculated
as(27):
T(Cj) = (EBCT/EBCTe1n) ~ T(C2)
i
where	T(Cj) ~ Throughput Producing Effluent Concentration, Cj
EBCTffl^p = Minimum Constant-Pattern EBCT
T(t2/ = Throughput Producing Effluent Concentration, C2
(C2 = Treatment Objective)
The constant-pattern HSD model was used to determine EBCT_^_,
T(C2), allowing direct calculation of T(Cj). Cj was determined from
T(Ci), again using the model results. Because constant-pattern adsorber
effluent is approximately equal to fractional capacity, fs was equated
to Cj/Cq. Figure 34 displays the manipulation of the constant-pattern
HSDM breakthrough profile to estimate fs and, thus, UR for two adsorbers
in series.

-------
22
The values of UR obtained from adsorption model simulations are
listed in Table 11. As expected, lower usage rates are obtained for
series adsorbers than single adsorbers, and longer EBCTs also reduce usage
rates. For all treatment conditions simulated, two 10-minute EBCT
adsorbers in series provided the optimum (lowest) URs. The relationships
between usage rate, adsorber configuration and EBCT, at three different
influent DBCP concentrations, are shown in Figures 35-37. Note that UR is
¦ore sensitive to treatment objective for single adsorbers than for series
adsorbers.
A water treatment cost model developed by CI ark(3) was used to
estimate adsorption costs. The model was based upon the work of
Culp/Wesner/Culp and the EPA Drinking Water Research Division(25).
Estimates produced by this cost model are considered "preliminary" and,
hence, relatively low in accuracy. The treatment cost model has been used
to estimate costs for regulatory impact analysis of Safe Drinking Water
Act regulations. The information necessary to make estimates with the
cost model, in addition to UR predictions derived from the adsorption
model, is presented in Table 12. Handling the spent activated carbon is
an important aspect of adsorption system design, as regeneration can
account for one-half or more of the total treatment cost. Because system
GAC usage was well below 450 kg/day (1000 lb/day), a value considered to
be the lower limit for economical regeneration(29), on-site regeneration
of spent activated carbon was not considered to be feasible. Instead, it
was assumed that GAC would be replaced on a contract basis, a service
offered by several activated carbon manufacturers. Alternatively, spent
activated carbon could be landfilled or incinerated. In practice, the
"throw-away" alternative is often preferred by users of small quantities
of GAC(29).
The cost estimates for DBCP adsorption are presented in Table 13.
The relationships between cost and treatment objective, for different
adsorber configurations, are plotted in Figures 38-40. Single adsorbers
achieve lower treatment costs than do series adsorbers, for a given EBCT.
The most economical treatment is attained with single 5- or 7.5-minute
EBCT adsorbers. Figures 41-43 display treatment cost as a function of
EBCT. As initial DBCP concentration decreases, the economy offered by
short EBCTs becomes more pronounced. Figure 44, the final cost plot,
shows the relationship between influent DBCP concentration and treatment
cost, for the various adsorber configurations and the 0.2 ug/L treatment
objective. Single adsorbers are less expensive than series adsorbers,
based upon these cost estimates.	1
The sensitivity of estimates produced by the cost model was tested by
varying design capacity/average flow ratio, GAC usage rate, and the
Interest rate, and observing the resulting changes in model output. Cost
model sensitivity was tested for the single 5-minute EBCT adsorber
configuration, DBCP influent concentration of 5 ug/L and treatment
objective of 0.2 ug/L. Increasing the usage rate by 100% over the
adsorption model prediction increased treatment cost by 35%. From the
standpoint of developing treatment cost estimates, then, accuracy in the
usage rate is important but not critical. Changing the design
capacity/average flow ratio from 9/4 to 6/4 resulted in a decrease in cost

-------
23
of 18%. Apparently, it may be advantageous - in terns of cost - to design
adsorption systems with less oversizing to meet peak flow. Reducing
excess capacity significantly reduces the capital cost for adsorbers.
What is being neglected in this analysis is the impact of a shorter EBCT
at peak flow upon carbon usage rate, which may increase operating costs
and negate some of the savings achieved in capital costs. Investigating
the impact of excess adsorber capacity upon UR and cost should certainly
be considered in design, and would be a straightforward model
application. Finally, changing the interest rate from 8% to 4% lowered
the treatment cost (annual cost; capital cost amortized over 20 years) by
about 15%.
The adsorption model, in conjunction with a cost estimation model,
was used to examine the dependence of treatment cost upon two design
parameters, adsorber EBCT and configuration, for a range of expected
treatment conditions. The results Indicated that the optimum design for
most conditions was a single adsorber utilizing a short (5-minute) EBCT.
A similar modeling exercise could be used to evaluate other aspects of
system design. For example, optimum operation of parallel adsorbers
allows longer operation of individual adsorbers, while still maintaining a
given treatment objective. The treatment efficiency achieved by parallel
adsorbers has been considered by several investigators(31,64), and is
reportedly most effective for treatment objectives (Ce/CQ) greater
than 30%(26). Adsorption models can assist in optimizing other design
aspects, as well - including selection of GAC to provide least-cost
treatment. Also, models nay be used to investigate design and operating
strategies to maximize efficiency under more complex conditions involving
¦ulticomponent adsorbates, NOM interference with target contaminants, and
biological activity(8,42).
CONCLUSIONS
1.	This study investigated the adsorption of DBCP from organic-free water
and ground water onto activated carbon. Adsorption capacity was
determined by bottle-point isotherm experiments. DBCP adsorption capacity
of activated carbon was lower from ground water than from organic-free
water. DBCP adsorption from ground water was tested with both virgin
activated carbon and an activated carbon "preloaded" by natural organic
matter (NOM) carried in the ground water. Preloaded activated carbon had
a substantially lower capacity for DBCP than did virgin activated carbon.
2.	A mathematical adsorption model, the pore-surface diffusion (PSD)
model, was used for analysis of kinetic experimental data and for
performance prediction. The PSD model was able to fit batch and plug-flow
adsorber data, from which mass transfer parameters were determined. The
predominant mass transfer mechanism for DBCP adsorption onto GAC,
intraparticle diffusion, was not significantly affected by solution
(organic-free water or ground water) or activated carbon (virgin or
preloaded). A lumped-parameter modeling approach was used to predict DBCP
adsorption performance under simulated water-treatment conditions (ground
water and preloaded GAC). The modeling approach was verified by comparing
calibrated model predictions to data from an independent adsorber
experiment.

-------
24
3.	The usefulness of mass transfer parameter correlations as substitutes
for kinetic experiments was evaluated. Correlations are attractive
because they are a simple alternative to conducting kinetic experiments,
which are difficult to conduct and analyze as well as time consuming.
Several literature correlations for the film transfer coefficient provided
acceptable estimates at low Reynolds numbers typical of GAC adsorbers. A
correlation for the surface diffusion coefficient provided accurate
parameter values for DBCP adsorption from organic-free water, but
overestimated the values for ground water and preloaded-GAC adsorption. A
parallel mechanism, pore diffusion, nay control intraparticle mass
transfer at these adsorption conditions. If so, then the
surface/diffusion coefficients could not be accurately determined in
ground water and preloaded-GAC experiments.
4.	The calibrated PSD model was used to make performance predictions for
full-scale adsorption of DBCP under drinking-water treament conditions. A
variety of adsorber empty-bed contact time (EBCT) and staging
configurations were investigated for different treatment objectives.
Activated carbon usage rates were determined by the PSD model predictions
for each adsorber design and treatment objective. The lowest usage rates
were obtained with long-EBCT adsorbers-in-series. The performance
predictions were then input to a treatment cost model, to develop cost
estimates. This allowed for the selection of the least-cost design, a
single 5-minute EBCT adsorber.

-------
25
REFERENCES
1.	Burlinson, N.E., Lee, L.A. and D.H. Rosenblatt. "Kinetics and
Products of Hydrolysis of l,2-Dibromo-3-chloropropane." Environ.
Sci. Technol., 16:9:627 (September 1982).
2.	Canonie Envir. Services Corp. "Development of Treatment Measures
for Ground Water Restoration." Report for Occidental Chemical Company of
Lathrop, California (July 1981).
3.	Clark, R.M. "Optimizing 6AC Systems." Journal Envir. Engrg.,
109:1:139 (February 1983).
4.	Crittenden, J.C. and W.J. Weber, Jr. "Predictive Model for Design
of Fixed-Bed Adsorbers: Parameter Estimation and Model Development."
Journal of Envir. Engrg., 104:2:185 (April 1978).
5.	Crittenden, J.C. and W.J. Weber, Jr. "Predictive Model for Design
of Fixed-Bed Adsorbers: Single-Component Model Verification." Journal of
Envir. Engrg., 104:3:433 (June 1978).
6.	Crittenden, J.C. and W.J. Weber, Jr. "Model for Design of
Multicomponent Adsorption Systems." Journal of Envir. Engrg., 104:6:1175
(December 1978).	'
7.	Crittenden, J.C. "Mathematical Modeling of Fixed-Bed Adsorber
Dynamics - Single Component and Multicomponent." Ph.D. Dissertation,
University of Michigan (1976).
8.	Crittenden, J.C. and D.W. Hand. "Design Considerations for GAC
Treatment of Synthetic Organic Chemicals and TOC," from AWWA Seminar
Proceedings, Strategies for the Control of Trihalomethanes, at Las Vegas,
Nevada (June 5-9, 1983).
9.	Crittenden, J.C., et al. "Prediction of Multicomponent Adsorption
Equilibria Using Ideal Adsorbed Solution Theory;" Environ. Sci. Technol.,
19:11:1037 (November 1985).
10.	Crittenden, J.C., et al. "Design Considerations for GAC Treatment
of Organic Chemicals." Journal AWWA, 79:1:74 (January 1987).
11.	Crittenden, J.C., et al. "Design of Rapid Fixed-Bed Adsorption
Tests for Nonconstant Diffusivities." Journal of Envir. Engrg., 113:2:243
(April 1987).
12.	Cummins, M.C. "Removal of l,2-Dibromo-3-Chloropropane (DBCP) From
Contaminated Ground Water by Packed Column Air Stripping." U.S. EPA,
Technical Support Division, Cincinnati, Ohio (1989).
13.	Daniel, C., Wood, F.S. and J.W. Gorman. "Fitting Equations to
Data". Wiley-Interscience, New York (1971).
14.	"Development of Best Availible Technology Criteria: Concept
Outline." Report Prepared for Science and Technology Branch, Criteria and
Standards Division, Office of Drinking Water, EPA (August 1987,
Unpublished).

-------
26
15.	Dobrzelewski, M. "Correlations for the Determination of Surface
Diffusivities of Organic Chemicals Adsorbed onto Granular Activated
Carbon." M.S. Thesis, Michigan Tech. University (1985).
16.	Draper, N. and H. Smith. "Applied Regression Analysis." 2nd Ed.,
John Niley & Sons, New York (1981).
17.	Environmental Protection Agency. Federal Register, 43:28:5756
(February 9, 1978).
18.	Environmental Protection Agency. Federal Register, 53:219:46936
(November 13, 1985).
19.	Environmental Protection Agency. Federal Register, 54:97:22062
(May 22, 1989).
20.	Envir. Sci. and Engrg. Inc. "Evaluation of the Treatability of
Ethylene Dibromide and Dibromochloropropane by Activated Carbon and Packed
Column Air Stripping." Report for Department of Environmental Regulation,
State of Florida (1983).
21.	Erskine, D.B. and W.G. Schulinger. "Activated Carbon Processes
for Liquids." Chemical Engineering Progress, 67:11::41 (November 1971).
22.	Friedman, G. "Mathematical Modeling of Multicomponent Adsorption
in Batch and Fixed-Bed Reactors." M.S. Thesis, Michigan Tech. University
(1984).
23.	Gaston, J. Internal Memoranda, "Activated Carbon Filtration for
DBCP Removal." Department of Health Services, State of California (January
7 and July 7, 1981).
24.	Gnielinski, V. "Gleichungen zur Berechnung des Narme- und
Stoffaustausches in Durchstromten Ruhenden Kugelschuttungen bei Mittleren
und Grossen Pecletzahlen." VT Verfahrenstechnik, 12:6363 (June 1978).
25.	Gumerman, R.C., Culp, R.L. and S.P. Hansen. "Estimating Water
Treatment Costs: Volume 1. Summary." Municipal Environmental Research
Laboratory, EPA, Cincinnati, Ohio (August 1977).
26.	Hand, D.W., Crittenden, J.C. and Thacker, W.E. "User-Oriented
Batch Reactor Solutions to the Homogeneous Surface Diffusion Model."
Journal Envir. Engrg., 109:1:82 (February 1983).
i
27.	Hand, D.W., Crittenden, J.C. and W.E. Thacker. "Simplified Models
for Design of Fixed-Bed Adsorption Systems." Journal Envir. Engrg.,
110:2:440 (April 1984).
28.	"Hawaii Monitors Environmental Exposures." Health and Env.
Digest, 1:7:5 (August 1987).
29.	Hess, A.F. "Design and Cost Data for Existing VOC Treatment
Facilities," Presented at AWWA Research Division Seminar, Treatment
Processes for the Control of Synthetic Organic Chemicals, Kansas City,
Missouri (June 14, 1987).

-------
27
30.	Lee, H.C. "Humic Substance Removal by Activated Carbon." Ph.D.
Dissertation, University of Illinois, Urbana (1980).
31.	Lee, M.C., et al. "Design of Carbon Beds to Remove Hunic
Substances." Journal Envir. Engrg., 109:3:631 (June 1983).
32.	Liu, K.T. and W.J. Weber, Jr. "Characterization of Mass Transfer
Parameters for Adsorber Modeling and Design." Journal WPCF, 53:10:1541
(October 1981).
33.	Lorber, M. Internal EPA Memorandum, "Update on Ground Water
Monitoring Data Base." Exposure Assessment Branch, Office of Pesticides
and Toxic Substances, EPA (August 1987).
34.	Lykins, B.W., Jr., et al. "Granular Activated Carbon for Removing
Non-Trihalomethane Organics from Drinking Water." Drinking Water Research
Division, MERL, Office of Research and Development, EPA (September 1984,
Unpublished).
35.	McCreary, J.J. and V.L. Snoeyink. "Granular Activated Carbon in
Water Treatment." Journal AWWA, 69:8:437 (August 1977).
36.	McCreary, J.J. and V.L. Snoeyink. "Characterization and Activated
Carbon Adsorption of Several Humic Substances." Water Research, 14:151
(1980).
37.	Michelsen, M.L. "An Efficient General Purpose Method for the
Integration of Stiff Ordinary Differential Equations." Journal AIChE,
22:3:594 (May 1976).
38.	Narbaitz, R.M. and A. Benedek. "Least Cost Process Design for
Granular Activated Carbon Adsorbers." Journal WPCF, 55:10:1244 (October
1983).
39.	0'Dahl, T. "Occidental Chemical Company at Lathrop, California, A
Ground Water/Soil Contamination Problem and a Solution." National
Enforcement Investigation Center, EPA, Denver, Colorado (1985).
40.	Ohashi, H., et al. "Correlation of Liquid-Side Mass Transfer
Coefficient for Single Particle and Fixed Beds." Journal Chem. Eng. in
Japan, 14:6:433 (June 1981).
41.	Personal Communication from Richard Miltner (1986).
42.	Personal Communication from John Crittenden, Michigan Tech.
University (1987).
43.	Personal Communication from Tom Speth, DWRD (1987).
44.	Pirbazari, M., et al. "Pilot-Plant Investigation of the
Adsorption of 1,2-Dichloroethane and l,2-Dibromo-3-chloropropane From
Water Supplies." Proceedings of the Engineering Foundation Conference held
at Schloss Elmau, Bavaria, West Germany (May 6-11, 1983).

-------
28
45.	Randtke, S.J. and V.L. Snoeyink. "Evaluating GAC Adsorptive
Capacity." Journal AWWA, 75:8:407 (August 1983).
46.	Roberts, P.V., Cornel, P. and R.S. Summers. "External
Mass-Transfer Rate in Fixed-Bed Adsorption." Journal Envir. Engrg.,
111:6:891 (December 1985).
47.	Safe Drinking Water Act, 42 U.S.C. 300f, et seq. (1974).
48.	Safe Drinking Water Act Ammendments, Pub. L. 99-339, 100 Stat.
642 (1986).
49.	Schneider, R. "Bestimung der Effectiven Korndiffusion-
skoeffizienten Ds nach dem Overflachendiffusionsmodell fur Korn-und Pulver
Kohlen Unterschiedlicher Partikelgrobe." M.S. Thesis, University of
Karlsruhe, West Germany (1982).
50.	Selleck, R.E., et al. "Application of Air Stripping Technology
for the Removal of DBCP Residues in Community and Industrial Water
Supplies." Report to Occidental Chemical Company of Lathrop, California
(1983).
51.	Smith, E.H., Tseng, S.K. and W.J. Weber, Jr. "Modeling the
Adsorption of Target Compounds by GAC in the Presence of Background
Dissolved Organic Matter." Envir. Progress, 6:2:18 (February 1987).
52.	Speitel, G.E., Dovanantzis, K. and F.A. DiGiano. "Mathematical
Modeling of Bioregeneration in GAC Columns." Journal of Envir. Engrg.,
113:1:32 (February 1987).
53.	Sudo, Y., Misic, D.M. and M. Suzuki. "Concentration Dependence of
Effective Surface Diffusion Coefficients in Aqueous Phase Adsorption on
Activated Carbon." Chem. Eng. Sci., 33:9:1287 (September 1978).
54.	Summers, R.S. "Activated Carbon Adsorption of Humic Substances:
Effect of Molecular Size and Hererodispersity." Ph.D. Dissertation,
Stanford University (1986).
55.	Suzuki, M. and K. Kawazoe. "Effective Surface Diffusion
Coefficient of Volatile Organics on Active Carbon During Adsorption from
Aqueous Solution." Journal Chem. Eng. of Japan, 8:1:379 (January 1975).
56.	Symons, J.M. "A History of the Attempted Federal Regulation
Requiring GAC Adsorption for Water Treatment." Jour. AWWA, 75:8:34
(August 1984).
57.	Tien, C. "Bacterial Growth and Adsorption in a Granular Activated
Carbon Column," in Activated Carbon Adsorption of Organics from the
Aqueous Phase, V.2 (McGuire, M.J. and Suffet, I.H., Eds.) Ann Arbor
Science (1980).

-------
29
58.	Transcript of EPA Program Peer Review, "Granular Activated Carbon
Research vs. Regulatory Agenda Needs for Phase II Organic Compounds." Held
March 19-20, 1986 in Cincinnati, Ohio.
59.	Verschueren, K. "Handbook of Environmental Data on Organic
Chemicals," 2nd Ed.; Van Nostrand Reinhold Co., New York, 1983.
60.	Wang, C.K. "Dynamic Modeling of Fixed-Bed Adsorbers Using a
Micro-Adsorbent-Diameter-Depth (MADD) System." Ph.D. Dissertation,
University of Michigan (1986).
61.	Weber, W.J., Jr. and J.C. Crittenden. "MADAM I - A Numeric Method
for Design of Adsorption Systems." Jounal Water Pollution Control
Federation, 47:5:924 (May 1975).
62.	Weber, W.J., Jr., and M. Pirbazari. "Effectiveness of Activated
Carbon for Removal of Toxic and/or Carcinogenic Compounds from Water
Supplies." EPA Municipal Environmental Research Laboratory, Cincinnati,
Ohio (1980).
63.	Weber, W.J., Jr. "Evolution of a Technology." Journal Envir.
Engrg., 110:5:899 (October 1984).
64.	Westernark, M. "Kinetics of Activated Carbon Adsorption." Journal
WPCF, 47:4:704 (April 1975).
65.	Williamson, J.E., Bazaire, K.E. and C.J. Geankoplis.
"Liquid-Phase Mass Transfer at Low Reynolds Numbers." I&EC
66.	Zimmer, G. and H. Sontheimer. "Activated Carbon Adsorption of
Organic Pollutants in the Presence of Humic Substances." Preprinted
Extended Abstract, Presented at Division of Envir. Chem., ACS Meeting at
Denver, Colorado (April 5-10, 1987).

-------
30
TABLE 1
ENVIRONMENTAL PROPERTIES OF D8CP
Molecular Weight = 236
Vapor Pressure = 0.8 mm Hg «» 20° C)
Solubility = 1000 mg/1 (@20° C)
Henry's Constant (est.) = 0.005 Atm-m3/m3 (612° C)
Hydrolysis 1/2-Life = 141 years (© pH 7, 15° C)
(Source: Reference 1)
TABLE 2
FREUNDLICH ISOTHERM PARAMETERS FOR DBCP ADSORPTION
Water/Activated Carbon Cone. Range (ug/L) Kf* n Correlation Coef.(r)
OFW/Virgin F-400	0.9-60	7.21 0.483 0.98
GW/Virgin F-400	3-270	4.54 0.396 0.99
GM/Preloaded F-400	4-70	0.883 0.502 0.99
(* Kf in Units of ug/L for Cfit mg/gm for qe)
TABLE 3
EXPERIMENTAL CONDITIONS FOR DCBR
DCBR
M
r
vs
C0
Mater/Activated
Exper.
(mg)
(cm)
(cm/s)
(ug/L)
Carbon
1
48.5
.0107
33.2
78
OFW/Virgin
2
48.5
.0107
33.2
76.7
OFW/Virgin
3
21.1
.0107
33.2
11
OFW/Virgin
4
120
.0315
14.2
80
OFW/Virgin
5
90.7
.0315
35.1
78.3
GW/Virgin
6
182
.0315
35.1
85
GW/Preloaded
\

-------
31
TABLE	4
RESULTS OF DCBR	EXPERIMENTS
DCBR	1-Param. Fit:	2-Param. Fit:
Ds(c» /s) 95% CI	Os(cb /s) 95% CI kf(c«/s)
5.9E-11 (4.4-7.4)	7.5E-11 (4-16) 0.035
6.4E-11 (5.5-7.2)	9.3E-11 (7-14) 0.026
3.3E-11 (2.0-5.4)	3.1E-11 (1.3-) 0.1
4.5E-11 (3.9-5.2)	3.8E-11 (3.3-4.5) 0.25
4.5E-11 (4.4-5.3)
4.0E-11 (2.5-5.6)
Dp(cb /s) 95% CI)
.0077 430 3.7E-6 (2.3-5.3)
Exper.
St
Bi
1
.035
26
2
.035
24
3
.017
16
4
.0062
28
5
.0039
98
6
.0077
370
(Alternative
Fit
TABLE 5
HIGH-FLOW SBA EXPERIMENTAL CONDITIONS AND RESULTS
GAC Size t(cb) vs(cb/s) N(gm) L(cb) kf(ca/s) 95% CI
60/80 .0107 33.2 0.68 3.4 0.0765 (.070-.081)
20/40 .0315 14.3 21 8.5 0.0213 (.018-.024)
TABLE 6
LONG-TERM SBA EXPERIMENTAL CONDITIONS AND RESULTS
t(cb) vs(cn/s) M(mg) L(cb) C0(ug/L)
.0107 1.33 100 0.45 102-111
Ds(cb2/s) 95% CI
Surface Diffusion Fit: 4.0E-11 (2.4-6.9)
Dp(c«n2/s) 95% CI
Pore Diffusion Fit: 4.0E-6 (1.3-9.0)
kf(cB/s)	B^
0.01	42
kf(cn/s)	B^
0.01	42

-------
32
TABLE 7
DYNAMIC VERIFICATION COLUMN EXPERIMENTAL CONDITIONS
r(c*>) vs(c«/s) M(mg) L(cbi) EBCT(s) C0(ug/L) qe(mg/gm)
.0107 1.33 1400 6.35 4.8	78-111 10.74
TABLE 8
COMPARISON OF kf CORRELATION ESTIMATES
TO VALUES FROM SBA EXPERIMENTS
High-Flow SBA	Long-Term
60/80 20/40	SBA
Reynolds No. (Re)	180 340	8.2
Experimental
Value of kf (cn/s) .0765 .0213	.01
Nilliamson's kf	.052 .018	.014
Gneilinski's kf	.065 .034	.015
Ohashi's kf	.065 .030	.014
TABLE 9
COMPARISON OF Ds CORRELATION ESTIMATES TO
VALUES FROM KINETIC EXPERIMENTS
Exper. C0(ug/L) Dobrzelewski's Ds(cn*/s) Experimental Ds(cm2/s)
DCBR 1	78	6.48E-11	(4.5-8.4)	5.9E-11	(4.4-7.4)
DCBR 2	76.7	6.26E-11	(4.4-8.1)	6.4E-11	(5.5-7.2)
DCBR 3	11	2.29E-11	(1.6-3.0)	3.3E-11	(2.0-5.4)
DCBR 4	80	6.40E-11	(4.5-8.3)	4.5E-11	(3.9-5.2)
DCBR 5	78.3	15.1E-11	(lli12)	4.5E-11	(4.4-5.3)
DCBR 6	85	15.8E-11	(11-21)	4.0E-11	(2.5-5.6)
L-T SBA	105	17.5E-11	(12-23)	4.0E-11	(2.4-6.9)

-------
33
TABLE 10
ADSORPTION MODEL INPUT
FOR FULL-SCALE PERFORMANCE PREDICTION
Freundlich Isotherm - Kf=.883, n=0.502
Ds(cmz/s) -
Co=20 ug/L ... Ds=1.72E-ll (Bi=59)
5 ... 7.43E-12 (Bi=69)
2 ... 4.27E-12 (Bi=74)
kf=0.00316 cm/s
r =0.05129 en
vs=0.339 cb/s
L(cm) -
EBCT=5 lin ... L=102
7.5 ... 152
10 ... 203
TABLE 11
PREDICTED FULL-SCALE USAGE RATES
EBCT Cft Cp Single Adsorber Series Adsorbers
(Bin) (ug/L) (ug/L) UR (kg/ML)	UR (kg/ML)
7.5
10
20
1
13.0
7.75

0.2
14.7
8.97

0.1
15.4
****
5
1
5.60
3.55

0.2
7.32
4.27

0.1
7.77
4.67
2
1
2.24
2.02

0.2
4.42
2.58

0.1
4.86
2.81
20
1
9.83
6.29

0.2
10.6
6.71

0.1
10.9
****
5
1
4.46
3.11

0.2
5.43
3.36

0.1
5.64
3.50
2
1
1.98
1.85

0.2
3.35
2.10

0.1
3.59
2.22
20
1
8.42
5.78

0.2
8.89
5.92

0.1
8.99
****
5
1
3.93
2.87

0.2
4.58
3.01

0.1
4.71
3.09
2
1
1.88
1.77

0.2
2.85
1.91

0.1
3.01
1.94

-------
34
TABLE 12
TREATMENT COST MODEL INPUTS
Electricity : 0.082 Cents/kW-hr
Labor : $12/hr
GAC : $ 0.95/lb (Initial Charge)
$ 1.25/lb (Replacement Charge)
Capital Recovery Factor : 0.101852 (8%, 20 yrs)
Producer's Price Index : 294 (1985 ave)
Construction Cost Index : 399 (1986 ave)
Individual Adsorber Surface Area, Volume and GAC Inventory:
EBCT(min) A(nz) V(m3) Wt(kg)
5	10.5 10.7 5110
7.5	10.5 16.0 7650
10	10.5 21.3 10200
TABLE 13
TREATMENT COST ESTIMATES
EBCT CQ Ce	Treatment Cost (Cents/lOOOgal):
(min) (ug/L) (ug/L) Single Adsorber Series Adsorbers
7.5
10
20
1
27.6
35.4

0.2
29.4
36.7

0.1
30.1
****
5
1
19.9
31.0

0.2
21.7
31.8

0.1
22.1
32.2
2
1
16.4
29.4

0.2
18.6
30.0

0.1
19.1
30.3
20
1
26.5
38.2

0.2
27.3
38.7

0.1
27.5
****
5
1
20.8
34.9

0.2
21.9
35.2

0.1
22.1
35.4
2
1
18.3
33.6

0.2
19.7
33.9

0.1
19.9
34.0
20
1
26.8
41.4

0.2
27.3
41.6

0.1
27.4
****
5
1
22.1
38.4

0.2
22.8
38.5

0.1
23.0
38.6
2
1
20.0
37.2

0.2
21.0
37.4

0.1
21.2
37.4

-------
FIGURE 1
T1me-to-EquiIibrium Determination for DBCP Isotherms
9
g 1 » ¦ 		1 | t «	»- >	1	•	|	»¦ ¦«	~ > « ¦*	>¦ I • ' '
\
o»®
3 6)
C
o
~»
<0
t-
~»
c
c
o
c
o
U  > » <	» t ~ «	» I « » «	
0	20BH	4000	G000	B000	10000	12000
Elapse Time (minutes)

-------
FIGURE 2
DBCP Rdsorption Isotherms
Freundltch Isotherm Plots:
om

, CM
©
OffrtlMM
/ %




/
A/a

i t t—* ~ » ~'
1	10	100
Liquid Phase Concentration, ug/1
1000

-------
FIGURE 3
DBCP Rdsorption Isotherms
RepresentitIon of Isother* Uncertainty!
Regression 95X Confidence Bounds
102 Error Limits

1000
Liquid Phase Concentration, ug/1

-------
FIGURE 4
Differential Column Batch Reactor
Run 1
CI
Best-Tit Model Simulation!
(OFW Isotherm Parameters)
Os " 5.9E-11 cm*cm/tec.
kf • 0.07B5 cm/sec.
4» in
u ra
m
to
a ® ..
rvi
3500
3000
2500
1500
se Time (minutes)
1000
El
500

-------
FIGURE 5
Differentia! Column Batch Reactor
Run 2
a>
Bsst-nt Model Slmulttloni
(OFW Isotherm P«r
Ds " G.4E-I1 cmKcm/sec.
kf ¦ 0.0765 cm/sec.
CT)
3 ® ..
w 10
- ® ..
+> in
CD
O S
(M
6000
5000
4000
2000	3000
Elapse Time (minutes)

-------
FIGURE 6
Differential Column Batch Reactor
Run 3
(M
Boat-Fit Model Simulation!
(OFH Isotherm Parameters)
Os " 3.3E-I1 emttcm/sec.
kf " 0.0765 cm/sec.
o>
*»
L 10
o
(M - -
9000
7000
4000
e Time (minutes)
El

-------
FIGURE 7
Differential Column Batch Reactor
Run 4
CD
Beat-Fit Modal Simulation:
COFH Isotharn Parametara)
Da " 4.5E-II cm*cm/aae.
Kf ¦ 0.0213 cm/sec.
is.
w tO
o *¦
U (B
m
Q (S
® ..
30000
25000
20R00
0
5000
Elapse Time (minutes)

-------
FIGURE 8
Differential Column Batch Reactor
Run 5
Best-Fit Model Simulation!
(GW Isotherm Parameters)
Ds ¦ 4.5E-I1 cm*cm/see.
kf ¦ 0.0324 cm/sec.
00
(O
(J ffl
m
Q s
CM
o ..
0
10000
Elapse T1 me (minutes)

-------
FIGURE 9
ifferential Column Batch Reactor
Run G
Best-Fit Modal Simulation:
-------
FIGURE 10
Surface Diffusivity Vs. Model Error
DCBR Run 1
... 1 parameter
l.E-10
Surface Diffusion Coefficient, Ds (cm*cm/sec.)
-------
FIGURE 11
Surface Diffusivity Vs. Model Error
DCBR Run 2
7.0
Surface Diffusion Coefficient, Ds (cm*cm/sec.)
-------
FIGURE 12
Surface Di-f f us i v i ty Vs. Model Error
DCBR Run 3
e-ii
l.E-10
l.E-9
Surface Diffusion Coefficient, Ds (cm*cm/sec.)
-------
FIGURE 13
Surface Dif-Fusivity Vs. Model Error
DCBR Run 4
. 1 parameter
2 parameters
\

,007
1. E-10
I: e-11
Surface Diffusion Coefficient, Ds (cm*cm/sec.)
-------
FIGURE 14
Surface Dif-fusivity Vs. Model Error
DCBR Run 5
cr
o
— ®
iH 5.3
l.E-9
1.E-I0
Surface Diffusion Coefficient, Ds (cm*cm/sec.)
-------
FIGURE 15
Surface Diffusivity Vs. Model Error
DCBR Run B
l.E-lt
Surface Diffusion Coefficient, Ds (cm*cm/sec.)
-------
FIGURE 16
PC PK" K'¦ i'l *1
Summation of Squared ( rror f( ntnur Plrt
iir i < i
Centoui
e
o
W
I .*E-11
1.E-I0
Ds (cm? sec - I i
-------
FIGURE 17
DC PK K'ijn 5
Summation of Squared FVror Contour Plot
Ds < cm2 sec-) )
-------
FIGURE 18
Values of Ds Determined from Kinetic Experiments
Best Fit and 95^ Confidence Interval
« ~ «¦»>»«

DCBR 1
o DCBR 2
DCBR 3
DCBR 4
DCBR 5
DCBR 6
.. Long-tern SBR
~ ¦ * > * * i
10
Surface Df f -fus I v t ty, Ds * 1.0E+11 cm*cm/sec
-------
FIGURE 19
DBCP Rdsorption Isotherms
and Comparison to DCBR Equilibrium Capacity
o
«o
I I I II llll
4< ¦ I I I I I I M
.» » 4 M MM
« ¦ » ¦> I Mill
E
o»
\
O)
E
¦o
«
JO
c S
o —
w
TJ
tr
~»
c
3
O
E
a:
© new a
y, OI-frilnM
6 DCBR S

* t ******
*¦ t t * I » » *
. 1
|	10	100
Liquid Phase Concentration, ug/1
1000
-------
FIGURE 20
Differential Column Batch Reactor
Run G
s
m
oi
c ®..
o OB
4»
O 8
Data as Reclaved
a
OC Corrected
in
0
20000
25000
15000
Elapse Time (minutes)
-------
FIGURE 21
High-Flow SBR Breakthru Profiles
60/80 Mesh GflC
Be«t-Flt Model Simulations
kf - 0.0765
-©
©
-|	1	|	•	~	<	1	•	1-
012345678
Elapse Time (minutes)
-------
FIGURE 22
High-Flow SBR Breakthru Profiles
20/40 Mesh GRC
Best-Fit Model Simulation!
kf - 0.0213
JSt	Q.
6J I t » t » > • I f t | » I >	f"
0	1	2	3	4	5	6	7	8	9	10 11 12
Elapse Time (minutes)
-------
FIGURE 23
Short-Bed Adsorber Breakthrough Profile
s
(VI T
Clfluint Data
Br
o>
Beat-*ft P&ra Dtffuotoni
Op-4.0E-6
M-a.ei
a
«/ Burfaoi Dlffualont
Da-4.K-ll
kf"€.01

1250
0
1000
1500
250
750
500
Elapse Time (minutes)
-------
FIGURE 24
Long-Term SBfl
Summation of Squared Error Contour Plot
1;E-II
i.e-10
Ds (cm? see-!)
-------
FIGURE 25
Long-Term SBfl
Summation of Squared Error Contour Plot
i.e-5
Dp (cm? see-I)
-------
FIGURE 26
Verification Column Breakthrough Profile
<9
CM
a>
O ©
EMIwnt Data ©
4*
4> U>
m/1 Pur* Dlffutfoni
0p«3.7E-S
kf-0.01
U
teat-fit ¦/ Surfaca Diffusions
Da-4.0E-ll
hf-fl.ei
(M
25000
15000
0
Elapse Time (minutes)
-------
FIGURE 27
Verification Column Breakthrough Profile
Model Sensitivity to 58% Variation in Isotherm Constant, Kf
o
- so*
(g --
Ol
CO
> sex
CD
Q
rvj
0
15000
Elapse Time (minutes)
-------
FIGURE 28
Verification Column Breakthrough Profile
Model Sensitivity to 50/C Variation in Surface Diffusivity
f\» T
O)
8 ..
4* U>
®
rvi
25000
15000
0
Elapse T1 me (minutes)
-------
FIGURE 29
Verification Column Breakthrough Profile
Model Sensitivity to 50/£ Variation in Film Transfer
s
rvj t
19
® ..
CD
© o
*- ® ..
+> to
® ..
OQ
® ..
rvj
0
5000
20000
Elapse Time (minutes)
-------
FIGURE 30
Verification Column Breakthrough Profile
Model Sensitivity to Influent Concentration Profile
<9

c ta
*» ID
Hod#! fitautitton •/ Mocftflcd Influent -
1BX Rtduotton tn Cennntritign up is t473B Ktn.
® ..
OJ
0
10000
15000
Elapse Time (minutes)
-------
FIG 31
Verification Column Breakthrough
Three Model Simulations Based
Prof i 1 e
Upon
Different Calibration Conditions
is
rvj -r
(9
CB +
/OBCP/GM/Pr*1otrf«4 GftC Byimm °
DBCP/OvVlrgln QIC SyotM
DBCP/OTM/VIrglrt QflC
5000
10000	15000
Elapse Time (minutes)
20000
25000
-------
FIGURE 32
Correlation for Surface Diffusivity
Compared to Experimental DCBR Results
•>
o
-------
FIGURE 33
Model Prediction of Full-Scale Rdsorber Performance
m
s
en
3
m
c
o
4>
c
o
o
c
o
u
Simulation for Following Conditional
Flow Rita ¦ S gp»/fW
EBCT - 9 »1n.
Co • 5 u|/l
(M
CL
U
m
a
3.E+B
2.E+G
1.5E+G
l.E+6
Elapse Time (minutes)
-------
FI pilar 34
VISUALIZATION OF CONSTANT-PATTERN ADSORPTION
USED TO ESTIMATE USAGE RATE FOR SERIES ADSORBERS
Constant-Pattern
Brejktfmnkih Profit*
CfCo
Tteatme rt Objects, C2/Co
qfqo
First Acfcorbcr
- Second Adsorber
00
L2
L1
0.0
As the second cohmn effluent, C2, approaches the treatment objective, the first adsorber Is removed from
service. Under constant-pattern conditions, the activated carbon in the first adsorber reaches a fraction of
its isothetm capacity, fs, equal to the adsorbate concentration leaving the first adsorber at the tfcnne It was
removed, C1/Co
-------
FK . 35
GRC Usage Versus DBCP Treatment Objective
for Different Rdsorber Configurations
20 ug/1 DBCP Influent Concentration
<0..
yj		^Jingle 5-mln EBCT
<
Q (M
\
O)
J*
'	Single 7.5-mln EBCT
5) "	"" 1	—		Single 10-mln EBCT
f? a> --	'	.—Series 5-mln EBCT
WW
ID
n		Series 7.5-mln EBCT
a:		'	-	Series 10-mtn EBCT
(J
6) 4
1	»
Treatment Objective (ug/1 DBCP)
-------
FT I 36
GRC Usage Versus DBCP Treatment Objective
for Different Rdsorber Configurations
5 ug/1 DBCP Influent Concentration
CD
W
<
CO
\
O)
JC
o
ay
a
V)
3
O
(£
KD
to --
IM --
® -t
*^-J5 Ingle 5-mln EBCT
	Single 7.5-mln EBCT
	Single l0-»1n EBCT
		Series 5-mln EBCT
	Series 7,5-mln EBCT
	Series 10-mln EBCT
i	I
Treatment Objective (ug/l DBCP)
-------
Flw
-------
F. .E 38
Treatment Cost vs. DBCP Treatment Objective
Different Adsorber Configurations
20 ug/1 DBCP Influent Concentration
	
	.	Serjes 7.5-iwIn EBCT
.	Series 5-mfn EBCT
~ —.— Single 5-»1n EBCT
:—~—===r.=r— llRglt >?5"A?nE!iiT
.1	I
Treatment Objective (ug/1 DBCP)
-------
,#E 39
Treatment Cost vs. DBCP Treatment Objective
Different Adsorber Configurations
5 ug/1 DBCP Influent Concentration
-»—>¦ * i
	 Series !0-mln EBCT
— J2--			 Series 7.5-mln EBCT
cn
®
CO
O
\
w o
4* n
C
V
u
4>
o
o in
L> ni
4>
c
«
E
~»
I-
in
~ . 1
			Series 5-mln EBCT
	 Single 10-mln EBCT
m	'—			"	 				 Single 7.5-mln EBCT
® ®-.	" "	Single 5-mln EBCT
	»_
Treatment Objective (ug/1 DBCP)
-------
FIGURE 40
Treatment Cost vs. DBCP Treatment Objective
Different Adsorber Configurations
2 ug/1 DBCP Influent Concentration
» ¦ i * ¦ +¦
rvi
	—		—				 Serlet 10-mln EBCT
-jl m..
0 «*»
o» —			 Sor 1 es 7.5-mln EBCT
S
o
\
H S .					
~j n f				—¦	Series 5-mln EBCT
c
o
U
~»
w
o tn
U ni
~»
c
V
E
~»
0
w a					Single 10-mln EBCT
		— —				— Single 7.5-nln EBCT
— Single 5-mln EBCT
m	| . « ¦ » ¦ ¦ » >!«*~		
.1	t
Treatment Objective (ug/1 DBCP)
-------
FIGURE 41
Treatment Cost vs. Individual
20 ug/1 DBCP Influent Concentration, 0.2 ug/1
EBCT
Treatment Objective
in _
CD
0 *
cn
©
o
CD
*¦4
\
m in
4< rn
C
V
O
~»
M
O to
U n
~»
C
o
E
id
® in
«- rvi
Series Rdeorbers
Single Rdtorber
s
rvi
1—
-4—4-
5
-•—h
6
-»—i—I i « «—h
7	B
-+¦
9
10
11
12
EBCT (minutes)
-------
F1 42
Treatment Cost vs. Individual EBCT
5 ug/1 DBCP Influent Concentration, 0.2 ug/1 Treatment Objective
S —
f- in .
, id «
O)
CD
CO
\
(I s
4t PI
c
V
o
~»
M
o in
(J CM
4>
c
1)
E
~>
id
O eg
t- rvj
Series Rdsorbere
Single Rdeorber
-------
FIGURE 43
Treatment Cost vs. Individual EBCT
2 ug/1 DBCP Influent Concentration, 0.2 ug/J Treatment Objective
® *
— in
« «*»
at
®
Q
(9
s
M Q
+> P)
c
o
u
~»
M
o in
U nj
~»
C
V
6
*»
(0
© (9
rvi
in
Series Adsorbers
Single Adsorber
0
EBCT (minutes)
10
It
12
-------
r iiiURE 44
Treatment Cost vs. Influent DBCP Concentration
0.2 ug/1 Treatment Objective
~ i ~ * >•
Single 10-mln EBCT
Serlee 10-mln EBCT
Series 7.5-«tln EBCT
Series 5-mln EBCT
Single 5-mln EBCT
Single 7.5-mln EBCT
~ i—*-
> > > ~—«—»—
10
100
Influent DBCP Concentration (ug/1)
-------
APPENDIX
EXPERIMENTAL DATA
ISOTHERM EXPERIMENTS
I. Organic-free Water/Virgin F-400: (Experinent and Analysis Performed
by DWRD)
Equilibration Tine = 8, 9 and 13 days
Bottle Voluae - 250 ¦!
Initial
Final

Anount
Cone
Cone
AC Dose
Absorbed
(ua/1)
(uq/1)
(¦q/D
(¦a/am)
492.3
1.7
53.63
9.148
492.3
44.5
37.02
13.18
492.5
5.1
29.44
16.55
492.5
10.8
21.94
21.95
492.5
18.7
15.32
30.91
492.5
31.6
11.25
40.95
492.5
69.2
9.073
46.63
266.3
0.9
39.48
6.722
266.3
2.5
19.72
13.38
266.3
6.0
14.02
18.56
266.3
11.5
9.44
26.99
183.7
1.2
26.1
6.992
183.7
3.1
16.6
10.88
183.7
5.5
10
17.82
II. Ground Water/Virgin F-400:
Analytic Method: EPA 504
Analyst: Southwest Research Institute
QC Data = saae as for DCBR Run 5
Equilibration Tiae = 13 days
Bottle Voluse = 1 liter
Initial
Cone
(uq/P
366
366
366
366
229.5
229.5
229.5
229.5
Final

Amount
Cone
AC Dose
Absorbed
(ua/1)
(«g/U
(ma/am)
46
13.7
23.3
99.5
8.4
31.7
174
5.4
35.6
269
2.6
37.3
3
32.2
7.03
8
22.0
10.1
19.5
14.9
14.1
44
10.0
18.6
-------
Ground Water/Preloaded F-400:
Analytic Method = EPA 504
Analyst: Southwest Research Institute
QC Data:
Det	Actual QC
QC Cone Cone	*
(uo/1)	(uq/1)	lixor
2.1
2
5
1.98
2
1
10.2
10
2
Equilibration Tiae = 27 days
Bottle Volume = 1 liter
Initial
Cone
(ug/1)
202
202
202
202
202
202
202
202
202
Final

Anount
Cone
AC Dose
Absorbed
(ua/1)
(¦q/i)
(¦q/ffm)
11
60.6
3.15
3.5
117.6
1.69
15
60.6
3.09
33.5
34.6
4.87
60
19.8
7.17
57
20.2
7.25
125
12.3
6.26
127
12.8
5.86
69
18.7
7.11
-------
DCBR RUN 1
Analytic Method: EPA 502.1
Analyst: Dan Hautman (TSD)
QC Data: Std. DBCP Cone = 99.2 ug/1
Determined QC Cone = 102.2
Actual QC Cone ~ 101.0
% Error =1.2
Concentration Data:
Time
Sample
(min)
Conc(uq/l)
0
78
270
64
605
59
1215
54
1500
47
2093
46
2655
43
3030
39
Experimental Conditions :
Flow = 1 1/min
Volume = 44.4 1
Weight GAC = 48.5 mg
Column Dia = 8
Depth = 3 am
Particle Rad = 0.0107 cm (60/80)
Water : Hi 11i-Q
pH = 8.0
T = 26°C
-------
DCBR RUN 2
Analytic Method: EPA 502.1
Analyst: Dan Hautman
QC Data: (Not Available)
Concentration Data:
Time
Sample

(¦in)
Conc(uq/l)
Due
0
76

0
74

0
80
92.6
330
76
66
505
59
63
875
61
61
1415
57

1595
49
50
1770
45

1950
45

2850
41
41
3030
41

3205
34.3
41.7
3385
38.6
40.9
4300
33.5
38.1
4820
38.5
36
5790
39
41
Experimental Conditions:
Flow = 1 1/min
Volume = 47 1
Weight GAC = 48.5
Column Dia - 8 ann
Depth = 3 mm
Particle Rad = 0.0107 cm (60/80)
Water: Milli—Q
pH = 8.0
T = 25°C
-------
DCBR Run 3
Analytic Method: EPA 502.1
Analyst: Dan Hautman
QC Data: Standard DBCP Cone =9.9
Determined QC Cone = 9.7 & 8.8
Actual QC Cone = 10.1
X Error = 4.2 & 12.9
Concentration Data:
Time
Sample

(min)
Conc(ua/l)
Pup
0
10
6.9
0
11

270
6.3
8.5
583
9.2
8.8
1170
7.5
7.9
1505
7.1

2041
6.7
6.3
2790
6.7
6.7
3100
7.0

4680
4.9

6930
6.0

Experimental Conditions:
Flow = 1 l/«in
Volume = 46.7 1
Weight GAC = 21.1 mg
Column Dia = 8 nm
Depth = 1.5 nm
Particle Rad - 0.0107 cm (60/80)
Water: Milli-Q
T = 25°C
-------
DCBR Run 4
Analytic Method: EPA 502.1
Analyst: Dan Hautnan
IL	«L
99.2	49.5
Determined QC Cone - 105	50.7
Actual QC Cone = 101	50.5
% Error = 3*9	0**
QC Data: Standard DBCP Cone
Concentration Data:
Tine Sample
(win)
Conc(ua/l)
0
80
60
71
480
61
1410
52
2850
44
4305
44
5730
39
7365
35
10030
30
11490
28
12930
29
14350
27
15825
29
17820
28
20115
24
23010
22
Que
79
43
38
35
32
32
30
23
25
Experimental Conditions:
Flow = 1.5 1/min
Volume = 46.5 1
Weight GAC = 120 ng
Column Dia = 15 mn
Depth = 2 mn
Particle Rad = 0.0315 cm (20/40)
Water : Milli-Q
T = 25°C
-------
DCBR RUN 5
Analytic Method: EPA 504
Analyst: Southwest Research Institute
QC Data:
Actual
Cone
iliflZll
0.02
1.0
5
10
30
50
75
QC
Determined QC
Cone	%
tug/D	Error
0.02	0
1.1	10
5.4	B
10.4	4
29	3
48	4
77	3
Concentration:
Tine	Sample
iiinl	Conc(uo/l) Dud
0	77
0	80
0	78
250	76
934	70 73
1423	69
2665	67
4075	60 64
4440	63
5420	60
6692	59 60
7187	54
8425	57
9580	58 58
10087	58
11255	56
12453	56 0
12936	55
14465	55 56
16780	56 54
17262	52
18509	55
19660	53 46
21100	52
Experimental Conditions:
Flow « 2.0 1/ain
Volume = 46.25 1
Weight GAC = 90.7 mg
Coluun Dia - 11 mi
Depth = 2 ma
Particle Rad « 0.0315 ca (20/40)
Water: Ground Water
T = 26°C
-------
DCBR Run 6
Analytic Method: EPA 502.1
Analyst: Dan Hautnan
QC Data:
STD Cone Det. PC Cone	Actual PC Cone % Errpr
80 97	101	4
80 97	101	4
80 99	101	2
80 95	101	6
80 107	101	6
80 87	101	14
Concentration Data:
Tine Sample
(gin) Conc(uq/1)	fiyfi
0	84
0	83
0	88
320	87
1500	78	79
3312	81	69
5643	78
6199	75
7094	79	75
7642	82
8525	80
8967	79
10039	76
10177	78	79
11403	74
11947	77
13460	68	66
15725	65
16287	61
17177	61
17596	62
18600	67
19157	69
20044	67
20570	70
21680	65	61
Experimental Conditions:
Flow = 21 l/«1n
Volume = 46.2 1
Weight = GAC 182.2 mg
Column Dia = 11
Depth = 4 Bin
Particle Rad = 0.0315 cm (Preloaded 20/40)
Water = Ground Water
T = 26°C
-------
HIGH-FLOW SBA EXPERIMENTS
Analytic Method: EPA 504
Analyst: Dan Hautman
QC Data:
Det QC Actual QC
Cone	Cone	*
(uq/1)	(ug/1)	Error
56	50	12
65	50	30
Experimental Data:
I. 60/80 - mesh 6AC
Time
(min)
Sample
Conc(ua/l)
0
107
0
116
0
103
1
29
2
35
3
31
4
38
6
36
8
41
10
36
Flow =
1 1/min
Weight 6AC = 680 mg
Column Dia - 8 an
Depth = 3.4 cm
Water: Milli-Q
T — 26 5°C
II. 20/40 - mesh GAC
Time
Sample
(min)
Cone(uq/1)
0
88
0
97
0
95
1
33
2
36
3
41
4
41
6
38
7
39
8
43
Flow =
4.2 1/min
Weight =21.0 grams
Column Dia - 2.5 cm
Depth = 8.5 cm
Water: Milli-Q
T = 25°C
-------
LONG-TERM SBA & DVC
Analytic Method: EPA 504
Analyst: Southwest Research Institute
QC Data:
Det QC
Conc(ug/l)
Actual QC
Conc(uq/l)
%
Error
2.1
10.2
1.98
2
10
2
5
2
1
Experimental Data:
Tine
Influent
(win)
Conc(uq/l)
0
44
1

2.5

5

15
102
30

60

120
111
180

240

300

360
108
780

1380
106
1785

2805
105
4190

4920

5900
105
6975

7860

8730
85
8735
55
9800

10285

11265
85
12090

13285

14485
85
15445

15920
95
17815

19910
78
*Note:
Analysis Performed
Dud
58.5
69
SBA Eff
Conc(ug/l)
74
85
87
72
97
95
100
101
102
102
104
105
105
Qua
DVC Eff
Conc(uq/l)
90
(new infl)
(solution)
(new infl)
(solution)
0.57
23*
14
51,
45
61
78
77
62
85
85
80
78
80
80
78
100
80
by Dan Hautnan
-------
Variable	SBA	BV£
Col. Dia =	8 Bin	8 im
Length =	4.5 wn	6.35 cb
AC Weight =	100 ag	1.402 gn
Plow =	40 ¦l/«rin	40 al/Bin
Mater =	6round	Water
T =	25°C	Water
-------