United States
Environmental Protection
Agency
Risk Reduction
Engineering Laboratory
Cincinnati, OH 45268
Research and Development
EPA/600/S2-91/048   Dec. 1991
Project  Summary
 Evaluation  of  the  MIDDAS
 System  for  Designing  GAC
 Adsorbers
Walter J. Weber, Jr., Margaret C. Carter, Kevin P. Olmstead, and Lynn E. Katz
  The micro-diaimster-depth-adsorptlon
system (MIDDAS) was evaluated for its
usefulness In determining equilibrium
parameters for adsorption In granular
activated carbon (GAC) systems.  The
system employs a column configura-
tion for determining such parameters,
rather than the traditional completely
mixed batch reactor (CMBR) configura-
tion. The equilibrium results were em-
ployed in the homogeneous surface dif-
fusion  model (HSDM) in  conjunction
with the short bod adsorber (SBA) tech-
nique to determine rate parameters for
trichloroethylena (TCE) adsorption on
GAC in both single-solute and  more
complex systems. The results of these
studies indicated  that the equilibrium
capacity for TCE adsorption from dis-
tilled water was lower to a statistically
significantly extent when determined by
the MIDDAS technique than when de-
termined by the CMBR technique. The
rate parameter associated with trans-
port within the adsorption particle (Dg,
the surface diffusion coefficient in the
HSDM) was significantly dependent on
the selected isotherm capacity param-
eter, whereas the rate parameter asso-
ciated with external mass transfer (k(,
the film transfer coefficient) was not
significantly affacted by changes In the
Isotherm.  These calibrated rate param-
eters were used in subsequent verifi-
cation  studies with deep adsorption
beds; the results  Indicated that rate
parameters  determined using the
MIDDAS technique tended to provide
better predictions of deep bed behav-
ior than did those determined with the
CMBR methodology.
  The MIDDAS methodology was also
employed In determining parameters for
both a bl-solute system Involving TCE
and p-dlchlorobenzene (DCB) and for a
system Involving TCE In a background
of naturally occurring  organic matter.
In the bi-solute case,  the equilibrium
interactions between the two  adsor-
bates were evaluated using the ideal
adsorbed solution theory (IAST), which
generally described the data well.
Model predictions resulting from the
parameters obtained with the  use of
these techniques were, however, not
as good, with the predictions for DCB
generally being superior  to those for
TCE.  In the study involving the pres-
ence of background organic matter,
TCE was added to Huron River water
before use in the MIDDAS system. The
predictions resulting from the param-
eters obtained in this case were poor,
indicating the possibility of a  depen-
dence of adsorption capacity on adsor-
bent particle  size or on other factors
not  accounted for in the MIDDAS ap-
proach.
  This Project Summary was developed
by ERA'S Risk Reduction Engineering
Laboratory, Cincinnati, OH, to announce
key  findings  of the research project
that Is fully documented In a separate
report of the same  title  (see Project
Report ordering Information at back).

Introduction and Background
  GAC treatment technology has been
designated as the best available technol-
ogy for removal of synthetic organic chemi-
cals  (SOCs) from contaminated  water.
While GAC  treatment  has indeed proven
                                                Printed on Recycled Paper

-------
to bo an excellent option for removal of a
broad range of SQCs commonly found in
raw water sources, difficulties still remain
with  regard to design of technically and
economically feasible systems.
  Fixed-bed (or column) adsorbers repre-
sent the most practical category of reactor
configuration for GAC treatment systems.
The  objective  of GAC system design in
such instances is to  optimize empty bed
contact time,  hydraulic surface loading,
and  system configuration  (i.e.,  series or
parallel column operation) to yield maxi-
mum utilization GAC adsorption capacity
white meeting specific treatment  objec-
tives. Various approaches have been de-
veloped to aid in the design process; com-
puter-based mathematical models offer the
dual advantage of furnishing the design
engineer with  important system informa-
tion, as well as providing  an avenue for
Indopth investigation of factors affecting
the adsorption process.
  The HSDM  is a predictive model that
has  been validated  over  an extensive
range of conditions. It incorporates math-
ematical descriptions of the major physi-
cochemlcal mechanisms  recognized  to
occur in fixed-bed systems: namely, axial
flow  with dispersion, local equilibria at the
surface of the particle, mass  transfer re-
sistance across a hydrodynamic boundary
layer  surrounding  the  particle, and
Intraparticle diffusion  along pore surfaces
within the particle.   Input parameters to
the HSDM consist of  phenomenological
rate  and equilibrium coefficients.  A vari-
ety of multi-parameter isotherm  models
generally describe GAC equilibria well. To
calibrate equilibrium models, two or more
characteristic constants must be  deter-
mined;  in  the case  of the widely used
semi-empirical Freundfch model, the pa-
rameters are an equilibrium capacity pa-
rameter, Kp, and  a parameter, n, which
relates to the magnitude  of  the  driving
force for adsorption.   Rate processes in
the HSDM, modeled as two resistances in
series, are characterized by an external
mass  transfer coefficient, k(, and an
Intraparticle surface diffusion constant, Ds.
  Accurate estimation  of  model  param-
eters is critical to reliably predict perfor-
mance. Isotherm parameters such as KL
and n are traditionally obtained from CMBR
equilibrium data.  Values for kf are typi-
cally determined from mass transfer cor-
relations derived from experimental data
for other systems (so-called "literature" cor-
relations),  and values  for  Df are com-
monly obtained by fitting CMBR rate data
with the use of an appropriate mathemati-
cal model. A number of  difficulties and
potential errors are associated with these
parameter estimation techniques.  Mass
transfer correlations are usually developed
for systems substantially different from the
system to which they are applied.  Fur-
thermore, the hydrodynamic and contami-
nant removal characteristics of CMBR rate
systems  differ dramatically from those of
the fixed-bed  reactors  used  in practical
GAC system designs.  It  is impossible to
predict which solutes or classes of solutes
may be most  influenced  by  reactor con-
figuration from first principles. If capacity
measurements are in fact strongly depen-
dent on reactor configuration for a particu-
lar GAC  application, then use of CMBR-
derived parameters to attempt simulation
of column configurations  may lead to an
inaccurate description  of adsorptive be-
havior of solution components.
  The above discussion delineates one of
the major difficulties associated with the
use of mathematical models for GAC sys-
tem  design: namely, accurate parameter
determination. The development of meth-
odologies to practically and economically
determine accurate input  parameters has
been a major focus of  research over the
past decade.  This research has  led to
development of a number of bench-scale
methodologies for determination of both
equilibrium  and  rate model coefficients.
The  MIDDAS technique employs a combi-
nation of the SBA technique and a modi-
fied version of a high-pressure minicolumn
method.  It was developed as a means to
provide greater accuracy in determining
model equilibrium parameters in  a system
with  the same  hydrodynamic attributes as
the full-scale adsorber, while doing so in a
manageable time frame.  The SBA tech-
nique was developed to  provide greater
accuracy in  estimating rate parameters.
The  MIDDAS  methodology allows  simul-
taneous  determination of  external film
transfer and intraparticle surface diffusion
coefficients from the same set of experi-
mental data; this obviates  the  need for
literature correlations  and CMBR-based
rate studies and eliminates potential error-
compounding by mutual compensation of
individual errors in these coefficients dur-
ing parameter search/regression analysis.
The  application of this methodology to es-
timate mass transport parameters has
been demonstrated previously,  and the
approach offers substantial promise as a
means for developing bench-scale infor-
mation that can be used  to facilitate and
enhance full-scale system design proce-
dures.
  The overall  goal of this study was to
extend previous investigations of the ap-
plicability of  the MIDDAS  methodology to
a range of practical multiple-solute/back-
ground  water supply conditions for se-
lected SOCs. A number of specific objec-
tives relating to this aim were investigated
in the studies described here, including:
1) comparison of MIDDAS and CMBR
methodologies for determination of equi-
librium parameters; 2) evaluation of the
SBA methodology for estimation of mass
transport parameters; and 3) evaluation of
ideal adsorbed solution theory (IAST) in
conjunction  with  the  MIDDAS  isotherm
technique as  an  approach to modeling
multisolute systems.

Materials and  Methods

Materials
  The adsorbent  used in all experiments
was Filtrasorb 400* activated carbon
(Calgon Corp., Pittsburgh, PA). GAC par-
ticle size fractions  were obtained by crush-
ing and sieving carbon samples obtained
from one lot of  carbon.  The resulting
fractions were washed in distilled, deion-
ized water to remove fines,  dried over-
night at 105°C, and transferred to airtight
containers for storage. Carbon for imme-
diate use was dried again to remove any
moisture adsorbed in  storage and stored
in a desiccator.
  TCE was the primary target SOC em-
ployed in this work; DCB was employed
as a second  target solute in bi-solute in-
vestigations.   TCE and DCB were se-
lected because they have been designated
as U.S. Environmental Protection Agency
priority pollutants  by the EPA, have been
identified  in  contaminated surface and
groundwaters, are  relatively straightforward
to analyze, and represent a broad class of
compounds commonly found  in the envi-
ronment.
  Feed solutions of the target solutes were
prepared through direct injection of a high-
concentration stock solution of the target
solute dissolved in methanol into the back-
ground  water.  Experiments were  con-
ducted at a temperature of 250±2°C. All
samples were extracted immediately into
hexane  and analyzed for TCE or DCB by
packed  or capillary column gas chroma-
tography with electron capture detection.
  To obtain  high purity water, distilled-
deionized water was processed through a
Milli-Q  water system  (Millipore).  Huron
River water  (HRW),  collected from Argo
Park in Ann Arbor, was stored and refrig-
erated in 55-gal stainless-steel drums un-
til used. Approximately 100 mg/L sodium
azide was added to retard bacterial growth
* Mention of trade names or commercial products does
 not constitute endorsement or recommendation for
 use.

-------
and degradation of organic matter. Batch
reactor isotherm studies have shown that
sodium azide does not significantly affect
the adsorption equilibria of  target com-
pounds.  Before use, the HRW was fil-
tered through a 1.0 u. filter (Gelman Sci-
ences) to remove particulate matter.  No
attempt was made to control  pH.

Methods

Isotherm Parameters from CMBR
Data
   Individual 250-mL borosilicate  bottles
containing preweighed masses of 80/100
mesh activated carbon were filled with a
solution  of  the  designated  background
water spiked with  either TCE or DCB.
Initial solution concentrations of the target
compound and carbon masses were esti-
mated to achieve equilibrium results that
spanned the concentration range of inter-
est.  Filled bottles were sealed headspace-
free with the use of Teflon-lined septa and
allowed to equilibrate on a rotary tumbler
for 7 days.   Rate studies with TCE and
DCB showed that an equilibration time of
7 days was sufficient to achieve  equilib-
rium for the  carbon size used in this study.
Filled, sealed bottles without carbon were
also tumbled simultaneously with  the
sample bottles to assess volatility losses.
After equilibration,  samples  were taken
from the bottles containing carbon using a
gaslight syringe (Hamilton), filtered through
a prewashed glass fiber filter (Gelman Sci-
ences) in  a stainless-steel  filter holder
(Fischer Scientific) to remove any fines in
the sample, and analyzed for TCE or DCB.
Other studies indicated that TCE and DCB
 losses  onto the filters and  filter holder
were negligible.  The mass of solute on
the carbon at equilibrium, qo, was obtained
 by a mass balance on the closed system:

         
-------
 Tabifl.lsothormParamotersforTCEinMilli-QWaterwiththBUseofMIDDASandCMBRTechniques

       Technique               Kp*                 n
MIODAS
95% Conf. limits
CMBR
95% Conf. Limits
1.10
(0.846, 1.43)
1.94
(1.72, 2.20)
0.564
(0.511,0.616)
0.527
(0.498, 0.555)
     *  Based on q. In ng/mg and C, in jtg/L
 7«W» 2. Effect of TCE Concentration on Rate Parameters
     Ce(ng/L)
D,(x10locmz/sec)
95% Conf. Limits
 k,(x 103 cm/sec)
95% Conf. Limits
104.6
398
1052
0.70
(0.26,2.8)
1.6
(0.7,3.4)
6.6
(2.2,250)
21
(6.7,30)
18
(10,23)
11
(8.3.14)
anticipated influent concentration as pos-
sible.
Large Column Verification
Studies
  A set of large column adsorber data for
TCE in Milli-Q water on 30/40 particle size
GAC was obtained.  The average influent
concentration was approximately 1000 u.g/
L.  Rate parameters obtained using both
the CMBR-based and the MIDDAS-based
isotherm and the  SBA technique are re-
ported in Table 3. Both parameter sets
ware  used  in the homogeneous surface
diffusion model to predict the breakthrough
profile for these systems.   The predica-
tions along with experimental breakthrough
data are presented in Figure 3. Although
early time predictions fit the data reason-
able well, in the latter portions of the break-
through, the MIDDAS-derh/ed parameters
clearly demonstrate a better fit to the data
than do those based on the CMBR iso-

7>W«3. Ftito Parameters for Largo Column Data
                     therm.  The CMBR-based  predication
                     shows a breakthrough delayed from the
                     actual data, which is indicative of the larger
                     capacity given by the batch reactor iso-
                     therm.

                     Bl-solute Predictions
                       SBA data were obtained for the bi-sol-
                     ute system consisting of DCB and TCE in
                     Milli-Q water.  To calibrate the bi-solute
                     version of the HSDM model, the SBA data
                     obtained from this experiment were  used
                     in  conjunction with the IAST model  and
                     with both the single-solute MIDDAS  and
                     CMBR isotherm parameters for TCE and
                     DCB.  The calibrated bi-solute parameters
                     were  used to  predict deep bed  perfor-
                     mance for a bisolute MIDDAS experiment.
                     Figure 4 presents one set of data with the
                     accompanying predictions.  The data rep-
                     resent the experimental breakthrough pro-
                     files for influent concentrations of  TCE
                     and DCB of 754 and 2,332 u.g/L, respec-
                     tively.   As can be seen in Figure 4, the
     GAC size  isotherm
          k(cm/secx 103)
       (95% Conf. Limits)
     p.(cm2/secx 1010)
              Limits
30/40
30/40
MIDDAS
CMBR
3.9 (3.8,4.2)
4.0 (3.7,4.2)
8.7(7.3,11.0)
3.2(3.0,4.1)
 prediction for the  DCB breakthrough by
.the MIDDAS-based parameters is very
 good, whereas that for the CMBFt-based
 parameters  overpredicts adsorption of
 DCB. The model also nicely captures the
 shape of the TCE breakthrough and over-
 shoot for both sets of parameters, although
 both predicted curves are displaced slightly
 from  the experimental data.

 TCE In Huron River Water
   The geometric mean regression method
 was used to determine the Freundlich pa-
 rameters  for  TCE  in  HRW.  Both the
 MIDDAS and CMBR experimental  meth-
 odologies were employed, with trends in
 1C and n similar to those observed for the
 Milli-Q water.  SBA data were obtained for
 TCE  in HRW for the 30/40 particle sizes.
 The data were used with the HSDM model
 to determine rate parameters for both the
 MIDDAS and CMBR isotherms. The re-
 sults  of the model calibrations are given in
 Table 4.
   Predictions were made of TCIE large
 column data with the use of both MIDDAS
 and CMBR isotherm and rate data (Figure
 5).  As can  be seen,  neither  parameter
 set predicts the data well, with both pre-
 dictions greatly overpredicting removal of
 TCE. Because even the MIDDAS method
 cannot yield a good prediction, additional
 factors not accounted for by the MIDDAS
 approach  are active in this system.  A
 rough approximation of the capacity of the
 carbon bed used in the large column veri-
 fication study indicates a value 25% lower
 than would be predicted by the MIDDAS
 isotherm conducted in HRW.  Although
 rate parameter determinations do indicate
 some reduction in D. in the HRW from the
 Milli-Q case,  which  can cause some re-
 duction in apparent capacity because of a
 slowing of the kinetics of adsorption, it is
 unlikely that as large of a capacity  reduc-
 tion as was estimated could be precipi-
 tated  by only a moderate reduction in D,.
  The major difference  between the
 MIDDAS and large column systems is par-
 ticle size.   The MIDDAS methodology,
 which uses the  smaller  80/100 particle
 size to expedite determinations  of equilib-
 rium data, is  predicated on different par-
 ticle sizes  having the  same equilibrium
 capacity.  Although this assumption ap-
 pears to hold for the studies conducted in
 the Milli-Q  water, additional  factors that
 violate this requirement are clearly mani-
 fest in the HRW study.  Further study is
 necessary to determine if modification can
 be  made  to  the MIDDAS technique to
 enable accurate description of  large col-
 umn breakthrough data when background
 organic matter is present.

-------
                                                              Three-way
                                                              valve
                              Headspace
                              free cylinder
                              Teflon
                              plunger
                         Injection and
                         refill port
                     Magnetic stirrer
            Divert flow line
            to waste
                                                                      Stainless steel
                                                                      tubing (typ.)
Figure 1. Schematic of the MIDDAS system.
                              100
                               10-
                                                   10
100

o
-------
                       O
                             1.2-
                             1.0-
                             0.8-
                             0.6-1
                             0.4-
                             0.2-
                             0.0
                                            a  data
                                           	MIDDAS prediction
                                           — GMBR prediction
                                            —j—
                                             20
                             40
                                        —I—
                                         60
—i—
 80
100
                                                   Throughput (1000 bed volumes)
       .  Prediction of large column breakthrough data for TCE in Milli-Q water using MIDDAS-
         and CMBR-defived Isotherm and rate parameters for 30/40 particle size.
                     o
                     5
1 -
                                                                        ,Q   DOB Data
                                                                        •   TCE Data
                                                                       	 MIDDAS DCB Pred.
                                                                       	 MIDDAS TCE Pred.
                                                                       — • CMBR DCB Pred.
                                                                       --  CMBR TCE Pred.
                                         20         40         60         80
                                                     Throughput (1000 Bed Volumes)
                                                        100
               120
FJgur»4. Sample prediction of bi-solute data.

-------
      1.0
o
Q
      0.8-
      0.6-
      0.4-
      0.2-
               B  .data

             	MIDDAS prediction

             —  CMBR prediction
      0.0-9
                      20           40          60           80          100

                          Throughput (1000 bed volumes)
Figure 5.  Prediction of large column breakthrough data for TCB in HRW using MIDDAS and CMBR-
         derived parameters


Table 4. Rate Parameters for TCE in Huron River Water

                   k, (cm/sec x 103)  D  (cm2/sec x 1010)
     GAG size          isotherm       (95% Conf. Limits)   (95% Conf. Limits)
      30/40
      30/40
MIDDAS
 CMBR
4.0 (3.7.4.3)
4.0 (3.2,4.4)
 1.6(1.3,2.1)
0.80 (0.74,0.98)
Conclusions
  The equilibrium capacity for TCE in I
Q water determined in the MIDDAS sys-
tem was statistically  lower than that ob-
tained using the CMBR method, whereas
the n value for the MIDDAS tended to be
higher than that of the CMBR.  The impor-
tance of these differences were confirmed
in large column verification studies for the
30/40 and 16/20 particle sizes; the predic-
tions made with the use of parameters
based on the MIDDAS isotherm yielded
predictions superior to those from CMBR-
derived parameters.
  As with the adsorption from Milli-Q wa-
ter, the  capacity for TCE in HRW in the
MIDDAS  system  was significantly de-
pressed from that in  the CMBR determi-
nations.  Unlike the  Milli-Q water results,
however,  neither  MIDDAS-  nor CMBR-
based parameters provided adequate pre-
                   dictions of a large column verification
                   study, with overpredictions of removal in
                   both cases.  Although isotherm determi-
                   nations indicated a reduction in capacity
                   because of competitive substances in the
                   HRW as well as compounding reductions
                   because of the column  configuration of
                   the MIDDAS protocol, an approximate es-
                   timate of the capacity of the carbon bed
                   used in the large column verification study
                   indicates a value lower still than was pre-
                   dicted by the MIDDAS isotherm in HRW.
                   Rate parameter  determinations do indi-
                   cate a reduction  in D, in the HRW from
                   the Milli-Q  case; these determinations can
                   cause some reduction in  apparent capac-
                   ity because of a slowing of the kinetics of
                   adsorption. It is unlikely,  however, that as
                   large a capacity reduction  as  was esti-
                   mated  could be precipitated by only  a
                   moderate  reduction in D,.  One obvious
difference between  the two systems  is
GAG particle size.  The smaller 80/100
particle size employed in the MIDDAS sys-
tem is to provide rapid attainment of equi-
librium parameters while still maintaining
the hydrodynamic attributes of a column
system. A major assumption in the devel-
opment of the MIDDAS protocol was that
equilibrium  capacity is not particle  size
dependent.  Although  no dependency  of
equilibrium capacity on particle  size was
manifest in the Milli-Q  water studies, par-
ticle size apparently does affect the equi-
librium capacity of TCE in HRW.  Thus,
although estimates in the MIDDAS  sys-
tem do improve predictions somewhat over
the CMBR case, it is clear that achieving
similar hydrodynamics  is not necessarily
enough to  ensure estimate of  accurate
isotherm parameters and that other fac-
tors, such as particle  size and the pres-
ence of background organic matter, must
be addressed in further refinement of the
MIDDAS methodology.
  In the bi-solute studies, reductions  of
TCE adsorption capacity in the  presence
of DCB were captured qualitatively by
IAST.  Predictions for  DCB breakthrough
were quite acceptable; the MIDDAS-de-
rived parameters achieved better predic-
tions of the data than did the CMBR coun-
terparts. Predictions of deep bed adsorber
data for TCE were, however, only moder-
ately satisfactory as best, comparable re-
sults  were  achieved  by using  both
MIDDAS- and CMBR-based parameters.
  The research presented here has sought
to bring parameter estimation methods one
step closer to the reactors commonly used
in practice.  The results have proved in-
conclusive, however. In some cases, the
capacity estimations  of  the  MIDDAS
method were better than those  of the
CMBR  technique;  the MIDDAS method
did not  universally improve predictions of
both single-  and bi-solute data over those
from CMBR-derived parameters.  More-
over, although the SBA methodology itself
remains a viable and facile method of rate
parameter determination, the demonstrated
dependency of the SBA-D, values on ca-
pacity measurements  makes such deter-
minations susceptible to propagation  of
isotherm inaccuracies and  hence brings
the problem full-circle.  The nature of these
findings indicates that there are still many
research areas deserving  of pursuit with
regard to applying the MIDDAS  approach
to simulation and to design of  fixed-bed
GAC reactors.
  It is clear from this and others' works
that an a priori assessment of which com-
pounds will be most significantly affected
by reactor configuration is  not possible at
                                                                          •&U.S. GOVERNMENT PRINTING OFFICE: 1992 - 648-080/40117

-------
this Juncture.  Further work Is required to
elucidate the exact mechanisms that bring
about the reductions seen in column based
systems. Factors that may play an impor-
tant role in determining the sensitivity of a
compound's capacity to reactor configura-
tion  Include the  presence of other
 adsorbing species, molecular size, hydro-
 phobicity, polarity, functional groups, and
 carbon type.
   Because the capacity of GAC for TCE
 is dependent on  particle size for adsorp-
 tion from HRW, the MIDDAS methodology
 may have to be modified.  Further studies
are necessary to determine the mechanis-
tic causes of the observed effect.
  The full report was submitted in fulfill-
ment of Interagency  Agreement CR-
814135-01-0 by the University of  Michi-
gan  under  the sponsorship of the U.S.
Environmental Protection Agency.
   W. J. Weber, Jr., M.C. Carter, K.P. Olmstead, and L.E. Katz are with the University
     of Michigan, Ann Arbor, Ml 48109-2125.
   Thomas F. Spath is the EPA Project Officer (see below).
   The complete report, entitled "Evaluation of the MIDDAS System for Designing GAC
     Adsorbers,"(OrderNo.PB91-234617/AS;Cost:$19.00,subjecttochange)willbe
     available only from:
           National Technical Information Service
           5285 Port Royal Road
           Springfield, VA 22161
           Telephone: 703-487-4650
   The EPA Project Officer can be contacted at:
           Risk Reduction Engineering Laboratory
           U.S.  Environmental Protection Agency
           Cincinnati, OH 45268
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati, OH 45268
BULK RATE
POSTAGE & FEES PAID
EPA PERMIT NO. G-35
  Official Business
  Penalty for Private Use $300
  EPA/600/S2-91/048

-------