b INNOVATIONS IN TRADITIONAL TREATMENT

versus
GAC
sorption by granular activate
I
n efficient and cost-effective
or removing specific organic
rinated groundwaters. The
IC treatment, however, is now escalating
adsorbent studied is more eflicient than because of the increasing costs of handling, trans-
GAC in removing trichloroethene from a porting, and disposing of spent GAC. These costs are
especially high for small water utilities whose GAC
groundwater-as well as being more usage rates are not large enough to justify on-site
thermal reactivation or that are located a long distance
resistant to fouling by natural organic from regional reactivation facilities.
matter. In the late 197Os, synthetic carbonaceous adsor-
bents, developed* for treatment of potable water,
were extensively tested and found to be very promis-
ing for removing synthetic organic chemicals (SOCs)
David W. Hand, Jon A. Herlevich Jr., from contaminated groundwaters. A major advan-
David L. Perram, and John C. tage of these adsorbents over GAC was that they
Crittenden could be regenerated on-site using steam or by solvent
extraction methods. Unfortunately, large-scale com-
mercial production of these synthetic adsorbents was
not available at that time.
More recently, an improved version-t of the car-
bonaceous adsorbents was developed and is now
commercially available in large quantities. This adsor-
bent is an improvement over earlier versions because
it has a higher adsorption capacity for SOCs, is more
efficiently regenerated, and has better mechanical
stability properties.2 Consequently, the adsorbent
allows for smaller fixed beds and adsorbent invento-
ries than the earlier synthetic adsorbents or GAC.

64 JOURNAL AWWA Copyright (C) 1994 American Water Works Association

 ous-phase solutions of
reagent-grades TCE were
prepared by equilibrating
TCE with organic-free
water (OFW). The OFW
was tap water that was
distilled, deionized, and
passed through a GAC
fixed-bed column. The
test water used was tap
water, which is an un-
treated groundwater in
Houghton, Mich. Table 1
summarizes some of the
.‘ - ..; water characteristics.
. .
. . * *. ..* During the preloading
* . . **. I*. studies, the total organic
. .

l . .:**
.* l

. . carbon (TOC) in the tap

water ranged from 1 to 3
mg/L. During the RSSCT
In this article, the perf ce of the improved and pilot studies, the TOC was 2.0 f 0.5 mg/L. Table
synthetic adsorbent in removing trichloroethene 2 summarizes some of the properties of the adsor-
(TCE) from a groundwater containing natural bents as determined by the manufacturers.
organic matter (NOM) was compared with that of Analyses. TCE samples (1.8 mL) were analyzed
GAC.$ Adsorption equilibrium, kinetic, and pilot- by direct aqueous injection3 using a gas chromato-
plant experiments were used to compare the adsor- graph (GC),** which was equipped with an electron
bents. Methods for predicting the performance of capture detector. The GC column was 240 cm long
the synthetic adsorbent, including rapid small-scale and had a 2-mm diameter.++ Forty mL/min of
column tests (RSSCTs) and mathematical models, argon-methane (95:5 ratio) was used for the carrier
were developed and verified by comparing them gas. The initial oven temperature was maintained at
with pilot data. 120%. The injection port and detector temperatures
were maintained at 2 10
and 35O’ -C,respectively.
The performance of a synthetic adsorbent in removing Experimental appa-
trichloroethene (TCE) from a groundwater containing natural ratus and procedures.
organic matter (NOM) was compared with that of granular Laboratory and experi-
activated carbon (GAC) . Adsorption equilibrium, kinetic, and mental procedures used
pilot-plant experiments were used to compare the adsorbents. in this study have been
Adsorption isotherm studies showed that the synthetic adsorbent presented in the litera-
had three to four times more TCE capacity than GAC over a TCE ture. The isotherm pro-
concentration range of 10-2,000 ug/L. When preexposed for 24
weeks to a groundwater containing NOM, the synthetic adsorbent *By Rohm and Haas Co.,
showed an average 12 percent loss in capacity compared with the Philadelphia, Pa.
single-solute isotherm performed on the virgin adsorbent; GAC +Ambersorb 563, Rohm and
Haas Co., Philadelphia, Pa.
showed an average 35 percent loss in its virgin capacity for a *Filtrasorb F-400, Calgon
preexposed time of only 10 weeks. Rapid small-scale column tests Coarbon Corp., Pittsburgh, Pa.
SAldrich Chemical Co., Mil-
and model calculations predicted pilot-plant effluent profiles for waukee, Wis.
the synthetic adsorbent. The pilot-plant data confirmed that the **Model 5880A. Hewlett
synthetic adsorbent is more efficient than GAC in removing TCE. Packard, Mt. View, Calif.
+tPacked with 3% SP- 1500
on 80x loo-mesh Carbopak B,
Sup&o, Bellefonte, Pa.

Copyright (C) 1994 American Water Works Association

AUGUST 1994 65
 at a loading rate of 5 m/h.
Schematic of minicolumn apparatus Samples of the. adsorbents
were removed from the
columns after 2.5, 10, and 24
weeks of exposure time. The
samples were freeze-dried and
placed in a desiccator to
groudwater remove the water.
source The minicolumn proce-
dure used to determine the
effective intraparticle surface
diffusivities (0,) was described
by Liu and Weber,6 Roberts et
al,7 and Crittenden et a1.8 The
experimental apparatus was
constructed of glass, PTFE, and
stainless steel in order to min-
imize contamination. Glass
columns* of 1. l- and 1.5 -cm
ID by 30-cm length were used
for the 60x80- and 20x50-
mesh-size particles, respec-
tively. PTFE end caps were
seated with PTFE-wrapped
rubber O-rings on each end of
1 Drain )
the column in order to ensure
a proper seal. The adsorbent
1 PUjnP 1 was supported on a loo-mesh
stainless-steel screen.
Figure 1 is a flow diagram
of the minicolumn. A con-
,..w 4 . Schematic of RSSCTs and pilot plant centrated TCE feed solution
was stored in a headspace-free
100-L gas-sampling bag,
which was supported by a
water bath to minimize the
strain on its seams. Two lab
pumpst were used to meter
the concentrated TCE solu-
tion and Houghton tap water
to a 13-L mixing tank.
Another lab pump* was used
to pump the solution from the
mixing tank to the mini-
columns. The average TCE
concentration pumped to the

P
minicolumns was 300 pg/L.
13-L Two stainless-steel valves
mixing were placed before and after
tank
each column for sampling and
Pump Pump flow rate measurements.
I d ’ The pilot-plant and RSSCT
experiments were conducted
in a temperature-controlled
room at 13%. Glass column@
cedures used were presented by Sontheimer et al4 of 5.08- and l.l-cm ID were used for the pilot and
and Randtke and Snoeyink.5 The adsorbents were RSSCT studies, respectively. Channeling and wall
preexposed to NOM by placing samples of 20x50-
mesh-size synthetic adsorbent and 12x40-mesh GAC *Ace Glass, Vmeland, N.J.
tMode1 QSY-0 FMI, Fhd Metenng, Oyster Bay, N.Y
into 5.0-cm-diameter glass columns. The columns *Model QSY-2, Fluld Metrmrg, Oyster Bay, N Y
were continuously treated with Houghton tap water §Ace Glass, Vineland, N.J.

Copyright (C) 1994 American Water Works Association

66 JOURNAL AWWA
 effects were avoided because the ratio of column
diameter to particle diameter was ~50.~ Figure 2 Characteristics of Houghton tap water
shows the pilot and RSSCT experimental design. The
Value
concentrated TCE solution was metered from the
gas-sampling bag into a 13-L mixing tank where it was l-3 mg/L
mixed with the groundwater. The TCE-spiked ground- 40-50 mg/L as CaCO,
140-150 mg/L as CaCO,
water was then pumped to the columns. Three-way 6.0 f 0.2
valves were placed at the top and bottom of each col- 1 13%
umn to collect the influent and effluent samples,
respectively.
The RSSCT method was discussed in detail by * :
m 4 Comparison of single-solute TCE
Crittenden et [Link]~tr On the basis of the results of isotherms for the synthetic adsorbent
the kinetic studies, RSSCTs were performed, assum- and GAC
ing that the intraparticle mass transfer rate was lin-
early dependent on particle size. The RSSCTs, 60 x 80-
and 100 x l20-mesh size, were designed to simulate
the pilot column that had an empty bed contact time s Data for synthetic adsorbent
(EBCT) of 2.5 min. The 60 x 80- and 100 x 120-mesh - - Myers Isotherm fit
l
sizes of the synthetic adsorbent were obtained by
crushing and sieving the virgin 20 x 50-mesh size. The 105
EBCTs for the pilot-plant studies were 2.5 and 10.0
min. The nominal influent concentration of TCE for
the RSSCTs and pilot columns was about 2,500 pg/L.
A large influent TCE concentration was chosen to
shorten the length of time required to perform the I
pilot studies. 104
Liquid-Phase Concentration-pg/L
Results and discussion
Equilibrium studies. Adsorption equilibrium
studies compared the single-solute TCE isotherm
capacities at low liquid-phase concentrations for the exposed to a groundwater containing NOM for 2.5,
two types of adsorbent. Figure 3 shows that for liquid- 10, and 24 weeks. Figure 4 shows the impact of NOM
phase TCE concentrations ~2,000 pg/L, the synthetic exposure on the isotherm capacity for the two adsor-
adsorbent has three to four times more capacity for bents. After 10 weeks of NOM exposure, there was no
TCE than the GAC. For example, at a liquid-phase observed decrease in capacity for the synthetic adsor-
TCE concentration of 100 pg/L, the capacity of the bent compared with an average decrease of 35 f 5
synthetic adsorbent is about 80 mg/g compared with percent for the GAC. After 24 weeks of NOM expo-
only 20 mg/g for the GAC. For a liquid-phase TCE sure, the average reduction in the isotherm capacity
concentration of 1,000 pg/L, the capacity of the syn- for the synthetic adsorbent from its single-solute
thetic adsorbent is about 200 mg/g compared with isotherm was only about 12 + 8 percent. The results
60 mg/g for the GAC. However, at liquid-phase TCE demonstrate that for a groundwater containing a rel-
concentrations greater than about 30 mg/L, the GAC atively low TOC concentration, the synthetic adsor-
has a greater capacity than the synthetic adsorbent.2 bent is fairly resistant to fouling by NOM.
As shown in Figure 3, the isotherm for the syn- Kinetic studies. Minicolumn studies were per-
thetic adsorbent exhibits a significant amount of cur- formed to evaluate the adsorption kinetics of the
vature when plotted on a log-log scale, and the Myers 60 x 80- and 20 x 50-mesh-size synthetic adsor-
isotherm equation was used to describe this curva- bent. The fixed-bed models were compared with
ture.r2 The Myers isotherm takes on the following form: the minicolumn data to determine the intraparti-
cle mass transfer parameters4t6 Because it was not
C = Q/H exp (I< QP) (1) known whether surface diffusion, pore diffusion,
or both control the intraparticle mass transfer rate,
in which C is the liquid-phase solute concentration the plug flow fixed-bed pore diffusion model (PDM),
(&IL), Q is the solid-phase solute concentration surface diffusion model (SDM), and pore and surface
(n&f/g), H and I< are constants that depend on tem- diffusion model (PSDM)4,13 were compared with
perature, and P is a constant. The numerical values of the minicolumn data. Because the single-solute
H, I<, and P were determined to be 1,405 L/g, 1.15 isotherms for the synthetic adsorbent showed some
(glmiM)p, and 1.76 (dimensionless), respectively. curvature when plotted on a log-log scale, Eq 1 was
Because NOM in water can significantly affect the incorporated into the fixed-bed model to describe the
adsorbent usage rate by reducing the adsorbent capac- adsorption equilibrium. Figure 5 displays the best-
ity and kinetics,4 isotherms were performed on the fit PDM, SDM, and PSDM calculations as compared
synthetic adsorbent and GAC that had been pre- with the minicolumn data for the synthetic adsor-

Copyright (C) 1994 American Water Works Association AUGUST 1994 67

 bent. Film transfer coeffi-
@ Properties of the adsorbents cients were determined using
a correlation developed by
Propwty Synthetic Adsorbent SAC Gnielinski.14 Film transfer
BET surface area-ms/g 550 950 was not found to be impor-
Micropore volume-mL/g 0.23 0.48 tant because the rate of
Mesopore volume-mL/g 0.14 0.09
Macropore volume-mL/g 0.22 0.04 adsorption was controlled by
Ash content-percent co.05 6.3 intraparticle diffusion. The
Water adsorption at 95% relative humidity- -&% 0.12 0.44 pore diffusion coefficients
Particle size-mesh 20 x 50 12 x 40
0.53 0.47 that were used in the PSDM
and PDM calculations were
determined from the follow-
ing equation:

gp-- 4 ‘ JEffect of NOM on TCE isotherm for the synthetic

Dp = cpDLhp (2)
“ :;f,;; adsorbent and GAC
yj ,:,
I._.,..i ..,r’
.*.‘i.>i,‘
” :~.~ l
,<!a,I.~~~~*~r~ Single-solute data for synthetic adsorbent in which D, is the pore diffusivity
,.f:‘ p*$,;r >?~Tx 0 Data for synthetic adsorbent, 2.5-week NOM exposure
,: .’ 4:,&T;!
., I I (cmz/s); D, is the liquid diffusivity
: * {,; : 3’ . Data for synthetic adsorbent, IO-week NOM exposure
,,,’ 1 f~ Data for synthetic adsorbent, 24-week NOM exposure (cmz/s); 7p is the intraparticle tortuos-
.,“,i.: 2:
‘ .z::
:.‘ :.i$.:-:,
:. + Single-solute data for GAC ity (dimensionless); and l p is the particle
11,. 0 Data for GAG, lo-week NOM exposure porosity (dimensionless). In both the
PSDM and PDM calculations, 7 was
assumed to be 1.0, which yiel cfs the
b
maximum intraparticle pore flux. The
results show that both the PSDM and
the SDM describe the minicolumn data,
a&%*E~O and the PDM does not. If the tortuosity
10s
Pg6 of the particle was lowered to ~1, then
the PDM would fit the data. However, a
tortuosity value < 1 is physically impos-

q&P I
sible, and this implies that surface dif-
fusion that acts in parallel to pore dif-
fusion is important. Therefore the SDM
was appropriate for determining the
I effective intraparticle diffusivities.
102 103 IO4
Table 3 summarizes the effective
Liquid-Phase Concentration-pg/L
intraparticle diffusivities based on sur-
face diffusion (0,) for the 60 x SO- and 20
x 50-mesh-size synthetic adsorbent. The
,, .’ .- observed D, value obtained for the 60 x
-‘ a PSDM, SDM, and PDM calculations compared with
go-mesh size was lower than for the 20
minicolumn data for TCE on the synthetic adsorbent
x 50 mesh size, which suggests that it is
a function of particle size. Kinetic studies
1.2 on GAC have shown that the surface dif-
. 20 X 50 Minicolumn data Best-fit PSDM Best-fit SDM
q q
fusivity is dependent on particle size of
e Best-fit PDM
I, i.,’ ;; 0 1.0 the adsorbent.r5,‘ 6 The differences in the
$t, :: .‘ ,X Q?
u D, values may be due to two intraparti-
I ”
5, 0.8
cle diffusion domains within the adsor-
‘% bent, and changing the particle size does
* 6c not alter the diffusion path length in the
2o 0.6
slow diffusion domain. As a result, the
8 rate does not increase inversely with R2
3 0.4 with smaller-size adsorbents as would be
i! expected if D, did not change. Conse-
= 0.2 quently, the overall effective surface dif-
fusion coefficient must decrease with
I decreasing particle size.
10.0 15.0 20.0 25.0 30.0 35.0 40.0 Because the D, value for the 60 x 80
Bed Volumes Treated-X Id’ mesh decreased approximately linearly
compared with that for the 20 x 50-mesh
particle size, the 60 x 80- and 100 x 120-

Copyright (C) 1994 American Water Works Association

68 JOURNAL AWWA
 mesh-size RSSCTs were designed assuming D, was a ::**
Effective surface diffusivities
linear function of particle size.
’ determined from the synthetic adsol bent
Comparison of RSSCT and pilot studies. The ‘ : minicolumn experiments
RSSCT method was tested for the synthetic adsorbent
by comparing RSSCT data with 2.5-min-EBCT pilot- Ad- Mesh Size OS cma/s
plant data. Table 4 summarizes the physical parame-
ters for the RSSCTs and the pilot columns. Figure 6 20x50 1.37 x 10-11
'60 x 80 t4.74 x 10-Q
compares the RSSCT data with the pilot data plotted
in terms of litres of water treated per gram of adsor-
bent (specific throughput). The 60 x SO-mesh-size
RSSCT predicted the initial profile of the pilot data
reasonably well up to about 60 percent breakthrough. shown in Figure 6. Table 5 shows the integrated col-
Between 60 percent breakthrough and exhaustion, umn capacities obtained from the RSSCTs and the
the RSSCT predicted a specific throughput which was 2.5-min-EBCT pilot-plant data, and the predicted
10 to 30 percent less than that achieved by the pilot isotherm capacities. The isotherm capacities were cal-
p1an.t. The 100 x 120-mesh-size RSSCT predicted a culated using Eq 1 and the average influent concen-
specific throughput that was 20 to 30 percent less trations for each column. The capacities that were
than that observed for the pilot column. In addition, observed for the RSSCTs were less than for the pilot
the RSSCT profiles appear to be steeper than the pilot plants and those predicted by the isotherm. The pilot-
data. When NOM fouling occurs, SOC profiles become plant capacity is the same as the isotherm capacity,
steeper because of displacement.4 which demonstrates that the 20 x 50-mesh size is
Based on the results shown in Figure 6, the design not affected by NOM fouling for this contact time.
of the RSSCTs using the scaling equations with the For the RSSCTs, the capacities are reduced, which
surface diffusivities assumed to be linear-dependent may be caused by NOM fouling. It may be possible
with respect to particle size seemed reasonable. If the that crushing the synthetic adsorbent makes it some-
design of the RSSCTs had been based on the surface what amenable to NOM fouling. The capacity of the
diffusivities being constant with respect to particle 60 x 80-mesh-size RSSCT is 9.7 percent less than
size, the RSSCT profiles would have appeared much that of the pilot column, and the capacity of the 100
earlier than those of the pilot data. This is because x 120-mesh-size RSSCTs is 14.8 percent less.
the EBCTs for the constant diffusivity RSSCTs would Overall, the RSSCTs satisfactorily predicted the
be about 2.5 times less than those for the linear dif- pilot-plant data, and they can be used for prelimi-
fusivity RSSCTs. nary design and cost calculations. The comparisons of
The capacities of the RSSCTs and pilot columns the RSSCT and pilot data for the synthetic adsorbent
were compared with predicted isotherm capacities are as good or better than the comparisons that were
in an attempt to explain the differences in the profiles reported by Crittenden et all7 for GACs.

Physical parameters for the RSSCTs and pilot columns

SyntlmtkAdwnbmt
Parameter GAG P&t* RAC Pilot? P&t* PIletS RSSCT** RasCrtT
Adsorbent packing charactefistibs
Bulk density+/cti 0.429 0.4340 0.527 0,538 0.449 0.468
Particle density-g/cm3 0.8034 .a034 0.85 0.85 0.85 0.85
Bad porosity 0.466 0.460 0.379 0.367 0.472 0.450
Particte radius--cm 0.0513 0.0513 .024 .024 0.0065 0.0105
AdSofbent ma&-g 200.0 800.0 224.5 889.8 10.87 10.89
Bow specFicatiofbs
Column diameter-cm 5.08 5.08 5.08 5.08 1.1 1.1
Column ler@bcm 23.0 91.0 21.0 82.5 25.5 24.5
EBCT-min 2.9 11.6 2.48 9.90 0.82 1.26
V@n/h 4.77 4.83 5.08 5.00 18.8 11.6
Bed volum~ti 466.2 1.844.4 425.6 1,672.O 24.2 23.3
Flow rate-eM/mifl 161.1 163.1 169.8 175.4 29.7 18.4
Average Muent TCE
concentfation-pg./t 2,440.o 2,550.O 2,500.o 2,290.o 2,640.O 2.660.0
Columncapa&y-mg/g 74.0 67.0 216.0 190.4 184.0 195.0
?',?"(i":,'+?f*
$ ..&>, Pilot column with B nominal 2.5min EBCT
’ * tPilot column with a nominal [Link] EBCT
~,@Q# +p(@ c*)umn wi* 0 &ins1 2,[Link] fZ@aT
i*::@& pilot column with a nominal lCL&min EBCT
* * *lOO x IZOmesh RSSCT simulsting the [Link] EBCT pilot column
~~,~~df” 1 ++a0 x -&J Rmr simul&iflg *e [Link] iTB(Jf pilot column
‘ ;,-,;:; >‘
.l’ “ ’
; ,, : _! :’I “ z‘ ;;?‘ ;~i~,t,’ 1

Copyright (C) 1994 American Water Works Association

AUGUST 1994 69
 Comparison of SDM cal-
Comparison of the RSSCT data with the pilot-plant data for the
culations and. pilot-plant
synthetic adsorbent
data. The predictive capabil-
0 Effuent of pilot (20 X 50 mesh)
ity of the SDM for the syn-
0 Effuent of RSSCT (80 X 80 mesh) thetic adsorbent was tested by
comparing model calculations
with the 2.5- and IO-min-
EBCT pilot-plant data. The
film transfer and surface dif-
fusion coefficients were
obtained from the correlation
developed by Gnielinski14 and
Table 3, respectively. Figure 7
shows that the model accu-
rately predicted the 2.5-min-
EBCT pilot data while slightly
overestimating the bed vol-
umes treated for the lO.O-
min-EBCT pilot data. The
kinetic parameters used for
the lO.O-min-EBCT calcula-
tion described the shape of
Specific Throughput--L wafer treated/g adsorbent the effluent profile. But the
Pilot (20 X 50) EBCT-2.50 min; RSSCT (60 X 80) EBCT-1.26 min; RSSCT column capacity was overes-
(100 X 120) EBCT-0.81 min timated by 12 percent using
the single-solute isotherm,
and this reduction in capac-
QURB 76 Comparison of the SDM prediction with the 25 and lD-min-EBCT pilot- ity was also observed in the
:.: plant data for the synthetic adsorbent
24-week preloaded isotherm.
The fact that only 12 percent
- - - Model prediction, IO-min EBCT 0 Pilot-plant data, IO-min EBCT of the capacity was lost for the
- Model prediction, 25min EBCT 0 Pilot-plant data, 2.5min EBCT lo-min-EBCT pilot column
and for 43 weeks of operation
shows the synthetic adsorbent
is resistant to fouling.
For Houghton groundwa-
ter with relatively low back-
ground TOC concentrations,
the SDM can be used to pre-
dict the performance of the
synthetic adsorbent. How-
ever, further testing of the
model is required to deter-

1:
c-/10.0
mine the limits of its ability

0.0
to predict adsorber perfor-
mance .
Synthetic adsorbent vep
20.0 30.0 40.0 A.0 6d.0 7d.O
sus GAC. A comparison of
the pilot-plant data for the
Bed Volumes Treated-X 18
two adsorbents showed that
the synthetic adsorbent is
more efficient than the GAC
in removing TCE from a
groundwater with a low-TOC
Wegrated Column PredIcted lsotbeml AvMage Influent
cencestratlon background. As shown in
Figure 8, the 2.5-min-EBCT
RSSCT (loo x 120) 2,640
synthetic adsorbent column
RSSCT (60 x 80) lasted about five times longer
Pilot, 2.6-min EBCT to initial breakthrough than
Pilot, l@min EBCT
the 2.5-min-EBCT GAC col-
umn did. The lo-min-EBCT

70 JOURNAL AWWA
Copyright (C) 1994 American Water Works Association
 GAC column only lasted 30
percent longer before initial ime to breakthrough for the 25 and lPmin-EBCT pilot columns for
TCE on the synthetic adsorbent and GAC
breakthrough than the 2.5-
min-EBCT synthetic adsor-
bent column. Figure 9 com- 0 Effuent, GAC 2.5min EBCT 13 Effuent, synthetic adsorbent, 2.5-min EBCT
pares the 2.5- and lO.O-min 0 Effuent, GAC tO.O-min EBCT = Effuent, synthetic adsorbent, Iq.O-min EBCT
EBCTs for the two adsorbents ---- lnfluent data
by showing effluent profiles i
i
in terms of reduced TCE con- ::
centration and specific 3,000
throughput. The 2.5 -min-
EBCT synthetic adsorbent col-
2,400
umn treated about five and
three times more water per
gram of adsorbent than the 1,800
2.5- and lO.O-min-EBCT GAC
columns, respectively. In
terms of initial breakthrough,
the 10-min-EBCT synthetic
adsorbent column increased 600
the specific throughput by
about 20 percent compared 0
with the 2.5-min-EBCT col- 0 100 IS0 260 260 360 3Lo
umn. As shown in Tables 3
and 5, the integrated column
capacities showed that the
2.5- and lO.O-min-EBCT syn-
thetic adsorbent columns T m Specific throughput profiles for the 25 and lO-miwEBCT pilot
have about three times more ,,.,., ‘ , ,.,,,‘ I columns for TCE on the synthetic adsorbent and GAC
field capacity than the 2.5- y+)-A\$@
*&>.2,>: *& ,,*,,,..*’
1u
1)i,$s‘ ,b:
and lO.O-min-EBCT GAC 0 Effuent, GAC 2.5min EBCT q Effuent, synthetic adsorbent, 2.5-min EBCT
) ;Cy;i
columns. Table 5 shows the l Effuent, GAG lO.O-min EBCT l Effuent, synthetic adsorbent, lO.O-min EBCT
,“L\,b,.-i
_,;’ :: 3,ooo
integrated column capacity
for the 2.5-min-EBCT column
was the same as that pre-
dicted from the single-solute
isotherm. The integrated col-
umn capacity for the lO.O-
min-EBCT synthetic adsor-
bent column was about 12
percent less than would be
predicted from the single-
solute isotherm. The capacity
observed for the lO.O-min-
EBCT column can be pre-
dicted by the 24-week NOM
preloaded isotherm shown in
Figure 4. 0 20 40 60 80 100 120 140 160
Specific Throughput--L water treated/g adsorbent
Summary
Isotherm studies demon-
strated that over a TCE con-
centration range of lo-2,000
pg/L, the single-solute adsorption capacity of the syn- synthetic adsorbent did not change, whereas the TCE
thetic adsorbent was three to four times higher than isotherm capacity of GAC was reduced by 35 percent.
that of the GAC. TCE isotherms performed on the RSSCTs based on a linear dependence of the intra-
adsorbents that were preexposed to a groundwater particle diffusivity on particle size can be used to
containing 1-3 mg/L TOC showed that the synthetic obtain preliminary design information for evaluat-
adsorbent was more resistant to fouling by the NOM ing the cost of treatment with the synthetic adsorbent.
than the GAC. After 10 weeks of adsorbent exposure RSSCTs performed with the synthetic adsorbent were
to the groundwater, the TCE isotherm capacity of the able to predict the 2.5-min-EBCT pilot-plant data.

Copyright (C) 1994 American Water Works Association

AUGUST 1994 71
The time saved using the RSSCT rather than a pilot 6. LIU, P.K.T. 6 WEBER, W.J. JR. Characterization of
study is about 50 and 80 percent for the 60 x 80- Mass Transfer Parameters for Adsorber Modeling
and 100 x 120-mesh sizes, respectively. The capaci- and Design. Jour. WPCF, [Link] (1981).
ties that were observed for the 60 x 80- and 100 x 7. ROBERTS, P.V. ET AL. The Influence of Mass Trans-
120-mesh-size RSSCTs were slightly less than the fer on Solute Transport in Column Experiments
capacity observed for the pilot data. The 60 x 80- With an Aggregated Soil. Jour. of Contaminant
mesh-size RSSCT had 9.7 percent less capacity, and the Hydrol., 1:375 (1987).
100 x 120-mesh size had 14.8 percent less capacity. 8. CRITTENDEN, J.C. ET AL. Design Considerations for
The lower capacity observed for the RSSCTs may be GAC Treatment of Organic Chemicals. Jour.
caused by NOM fouling because crushing the virgin AWWA, [Link] (Jan. 1987).
20 x 50-mesh-size synthetic adsorbent to the particle 9. MARTIN, H. Low Peclet Number Particle-to-Fluid
size used in RSSCTs may make it more susceptible Heat and Mass Transfer in Packed Bed. Jour.
to NOM fouling. Chem. Engrg. Sci. [Link] (1978).
A minicolumn study combined with the SDM 10. CRITTENDEN, J.C.; BERRIGAN, J.K.; &- HAND, D.W.
predicted effluent profiles and usage rates for the Design of Rapid Small-Scale Adsorption Tests for
synthetic adsorbent for Houghton groundwater. More a Constant Diffusivity. Jour. WPCF, [Link]
testing is required to determine the limits of the (1986).
model’ s ability to predict performance. The incen- 11. CRITTENDEN, J.C. ET AL. Design of Rapid Fixed-
tive for this is the time savings involved. It takes Bed Adsorption Tests for Non-Constant Diffu-
only four days to perform a minicolumn study, com- sivities. Jour. Envir. Engrg. Div.-ASCE, [Link]
pared with a pilot study that takes a minimum of (1987).
120 and 300 days for the 2.5-and lO.O-min EBCTs, 12. JOSSENS, L. ET AL. Thermodynamics of Multi-
respectively. Solute Adsorption From Dilute Solutions. Chem.
Pilot-plant data showed the synthetic adsorbent is Engrg. Sci., [Link] (1978).
more efficient than GAC in removing TCE from a 13. FRIEDMAN, G. Mathematical Modeling of Multi-
groundwater. The 2.5-min-EBCT synthetic adsorbent component Adsorption in Batch and Fixed-Bed
column lasted five times longer to initial breakthrough Reactors. Master’ s thesis, Michigan Technol.
than the 2.5-min-EBCT GAC column, and its break- Univ., Houghton ( 1984).
through was slightly before the lo-min-EBCT GAC 14. GNIELINSKI, V. Gleichungen zur Berechnung dus
column. The synthetic adsorbent column had three Warme- und Stoffaustausches in durchstromten
times more capacity than the 2.5- and lO.O-min- ruhenden und grossen Pecletzahlen. VT Ver-
EBCT GAC columns. NOM fouling did not affect the fahrenstechnik, [Link] (June 1978).
2.5-min-EBCT synthetic adsorbent column. 15. BUCHHOLZ, H. fr KR~CKELS, W. Sorptionskinetik
geliister organischer Saubstanzen an Aktivkohlen.
Acknowledgment Verfuhrenstechnik, [Link] (1976).
The authors thank the Rohm and Haas Co. for its 16. FRICK, B. Unveroffentlichte Messungen. Bereich
technical assistance and financial support and the fur [Link], Univ. Karlsruhe, Germany
Environmental Engineering Center (EEC) at Michi- (1982).
gan Technological University for its financial support. 17. CRITTENDEN, J.C. ET AL. Predicting GAC Perfor-
The views expressed here are those of the authors mance With Rapid Small Scale Column Tests.
and do not necessarily reflect the views of the Rohm Jour. AWWA, [Link] (Jan. 1991).
and Haas Co. or the EEC. The mention of trade names
does not constitute endorsement for use by the About the authors: Duvid W
authors or the EEC. Hand is an assistant professor in the
Department of Civil and Environ-
References mental Engineering, Michigan Tech-
1. NEELY, J.W. &- ISACOFF, E.G. Carbonaceous Adsor- nological University (MTU), 1400
bents for the Treatment of Ground and Surface Waters. Townsend Dr., Houghton, MI 49931.
Marcel Dekker, New York (1st ed., 1982). He IS also associate director of the
2. PARKER, G.R. JR. 6 BORTKO, S.M. Groundwater Environmental Engineering Center
Remediation Using Ambersorb Adsorbents. Proc. at MTU. For the past 11 years, Hand’ s research has focused
Florida Envir. Chem. Conf., West Palm Beach, on the development of physical and chemical treatment
Fla., Oct. 30-Nov. 1, 1991. processesfor removing organics from water. He is a gradu-
3. GC Bull. 816A, Supelco, Bellefonte, Pa. ate of MTU with BS, MS, and PhD degrees. A member of
4. SONTHEIMER, H.; CRITTENDEN, J.C.; 6 SUMMERS, AWWA, Hand was a corecipient of the Engineering and
R.S. Activated Carbon for Water Treatment. Construction Division’ s best paper award at the 1988 annual
DVGW-Forschungstelle, Engler-Bunte-Institut, conference. Jon A. Herlevich Jr. is a chemical engineering stu-
Universtat Karlsruhe, Germany (2nd ed., 1988). dent, David L. Perram is a research scientist in the Envi-
5. RANDTKE, S.J. 6 SNOEYINK, V.L. Evaluation of GAC ronmental Engineering Center, and John C. Crittenden is
Adsorptive Capacity. Jour. AWWA, [Link] (Aug. presidential professor in civil and environmental engineer-
1983). ing-all at MTU.

72 JOURNAL AWWA Copyright (C) 1994 American Water Works Association