&EPA
                            United States
                            Environmental Protection
                            Agency
                                National Risk Management
                                Research Laboratory
                                Ada, OK 74820
                            Research and Development
                                EPA/600/S-99/001
February 1999
ENVIRONMENTAL
RESEARCH   BRIEF
            Colloid Mobilization and Transport in Contaminant Plumes:
            Field Experiments, Laboratory Experiments, and Modeling
                       Joseph N. Ryan1, Rebecca A. Ard1, Robin D. Magelky1,
                       Menachem Elimelech2, Ning Sun2, and Ne-Zheng Sun2
Abstract
    The major hypothesis driving this research, that the
transport of colloids in a contaminant plume is limited by the
advance of the chemical agent causing colloid mobilization,
was tested by (1) examining the dependence of colloid
transport and mobilization on chemical perturbations, (2)
assessing the relative transport of mobilized colloids and
the chemicals that caused their mobilization, and (3)
developing a colloid transport model that would begin to
describe these effects. Through field tests, laboratory
experiments, and model development, we made significant
advances toward the testing of the hypothesis. The field
tests, conducted in the uncontaminated and contaminated
zones of a ferric oxyhydroxide-coated quartz sand aquifer,
showed in almost all cases that colloids will not advance
ahead of the plume that caused their mobilization. The
laboratory experiments showed chemical perturbations that
cause increasingly repulsive conditions produced more
extensive and more rapid colloid release. In both the field
and laboratory experiments, good correlations wereobserved
between the surface properties of the colloid  and aquifer
grains and their transport and mobilization behavior. The
colloid transport model was developed to describe colloid
' Department of Civil, Environmental, and Architectural Engineering,
 University of Colorado, Boulder.
2 Department of Civil and Environmental Engineering University of
 California, Los Angeles.
                     transport in physically and geochemically heterogeneous
                     porous media similar to that encountered at the field site.
                     The model results showed that physical and geochemical
                     heterogeneities could resultin preferential flow of colloids in
                     layered porous media, while random distributions of the two
                     heterogeneities, especially the physical heterogeneity, lead
                     to a random behavior of colloid transport.
                     Introduction
                         Colloids have been implicated in the enhanced transport of
                     radionuclides and metals  in recent field studies and laboratory
                     experiments [Buddemeierand Hunt, 1988; McCarthy and Zachara,
                     1989; Dunnivant et al., 1992; Puls and Powell, 1992; Grolimund
                     etal., 1996]. Unfortunately, these studies have rarely delved into
                     the genesis, nature, and abundance of the colloids responsible
                     for the  enhanced  transport. We are relatively certain of the
                     processes governing the association of these contaminants with
                     colloids, but we  have  little knowledge of the potential for colloid
                     mobilization and subsequent transport in a given aquifer.
                         Colloid mobilization  is caused by chemical and physical
                     perturbations to  aquifer geochemistry and hydraulics [McCarthy
                     and Degueldre,  1993; Ryan and  Elimelech, 1996]. Chemical
                     perturbations of the type occurring in contaminant plumes are
                     capable of mobilizing large quantities of colloids. In particular, the
                     mobilizing effect of organic compounds like surfactants and
                     reductants is well known [Ryan and Gschwend, 1994; Allred and
                     Brown,  1994]. Colloid mobilization by physical perturbations is
                     generally limited to fracture flow and increases in groundwater
                     flow velocity induced by pumping [Degueldre etal., 1989; Pulset
                     al., 1992; Backhus et al., 1993].

-------
     In  contaminant plumes,  colloids are  mobilized and
transported with the groundwater.  If the advance  of the
contaminant plume is retarded, the colloids will attempt to move
ahead of the plume. When  the  colloids re-enter the pristine
groundwater, they will be redeposited. As the plume catches up,
the colloids will be remobilized and the cycle will begin again.
These simultaneous mobilization and deposition processes have
been observed  in natural analogs to contaminant plumes like
organic  matter-rich water infiltrating from a swamp [Ryan and
Gschwend, 1990] or fresh water advancing into salt water in a
coastal aquifer [Goldenberg et al., 1983]. The major hypothesis
of this research was that the transport of the colloids in  a
contaminant plume is limited by the advance of the chemical
agent causing colloid mobilization. To design experiments to test
this hypothesis, we set three overall objectives  for the project:
(1) examine the dependence of colloid transport and mobilization
on chemical perturbations, (2) assess the relative transport of
mobilized colloids and the chemicalsthat caused their mobilization,
and (3) develop a colloid transport model that would begin to
describe these  effects.  These objectives were met and the
hypothesis was tested by field experiments atthe U.S. Geological
Survey Cape Cod aquifer site, laboratory mobilization experiments,
and the  development of a colloid transport model emphasizing
the dynamics of colloid transport and the effects of heterogeneity.
Field Experiments
Purpose
     Two field experiments were conducted during the summers
of 1996  and 1997 at the U.S.  Geological Survey's  Toxic
Substances Hydrology  Research  Site on Cape  Cod,
Massachusetts. To address two of the overall objectives of the
project,  to  examine the  dependence  of  colloid transport and
mobilization on chemical perturbations and to assess the relative
transport of mobilized colloids and the chemicals that caused
their mobilization, the  following tests were conducted in the field:
•  assessment of the effects of chemical perturbations (elevated
   pH, surfactant concentration,  and  reductant concentration
   and decreased ionic strength) on the mobilization of natural
   colloids, synthetic colloids, and viruses
•  measurement of the rate of deposition of synthetic colloids
   and viruses
•  determination of the relative rates of migration of a plume of
   a colloid-mobilizing agent and the mobilized  colloids
In addition, we demonstrated the use of silica-coated metal oxide
tracer colloids  in  a field experiment unrelated  to the original
overall goals of the project.
     These experiments were conducted in the field (ratherthan
in the laboratory )forthree major reasons. First, we expect that the
extent of colloid mobilization from a sedimentwould increase with
disturbances to  the  sediment like sampling,  repacking into
columns, and  drying. Measurement  of the extent of colloid
mobilization in situ should provide a more realistic estimate of the
effect of chemical perturbations on colloid  mobilization. Second,
the Cape Cod site provided a unique opportunity to observe
colloid mobilization and transport in a relatively homogeneous
aquifer with a clear geochemical difference imposed by a plume
of secondarily-treated sewage; reproducing such a system in the
laboratory would be very difficult. Third, we hypothesized that
detection of the relative rate of migration of a colloid-mobilizing
agent (e.g., elevated  pH, surfactant) and the mobilized colloids
would require monitoring of  transport  of a distance of  a few
meters. The Cape Cod site provided well-instrumented arrays of
multi-level  samplers  for  injection  and  monitoring to test this
hypothesis.
Site Description
     The virus and silica colloid injections were conducted in the
surficial aquifer at the U.S. Geological Survey's Cape Cod Toxic
Waste Research Site nearthe Massachusetts Military Reservation
on Cape Cod, Massachusetts. The aquiferwas contaminated by
disposal of secondary sewage effluent onto rapid infiltration sand
beds for over 50  years [LeBlanc, 1984], creating a contaminant
plume characterized by low dissolved oxygen concentrations
and  elevated  pH,  specific  conductivity, and organic  carbon
concentrations (Table 1). Previously, the site has been used to
study the  transport  of groundwater tracers, metals,  nutrients,
detergents, microspheres, bacteria, protozoa, and viruses [Barber
et al., 1988; Harvey et al., 1989; 1995; Harvey and Garabedian,
1991; Smith etal., 1991; Hessetal., 1992; Kentetal., 1994; Bales
etal., 1995; Pieperetal., 1997]. The aquiferconsists of Pleistocene
glacial outwash deposits characterized by interbedded lenses of
sand and gravel.  The grains (average diameter 0.6 mm) consist
mainly of quartz  coated by ferric oxyhydroxides [Coston et al.,
1995]. The effective porosity is 0.39 and the average hydraulic
conductivity is 110 m d~1. The water table depth is between 6 m
and 7 m below the surface near the study site [LeBlanc et al.,
1991].
     Groundwater  at the site  is monitored by approximately
1,000 multi-level  samplers (MLSs), each consisting of a bundle
of polyethylene tubes threaded through a polyvinylchloride pipe
to 15 depths below the watertable at 25 cm depth intervals. In the
direction of groundwater flow, these MLSs are spaced at 1m to
2 m distances. For our studies, we  used three small arrays of
MLSs near the up-gradient end of the site (Figure 1).
Materials and Methods
     The following section highlights important aspects of the
materials used and methods followed during the field experiments.
Details are provided in Ard [1997] and Magelky [1998].
     Injections.  Two major sets  of  field  experiments were
conducted during the summers of 1996 [Ard, 1997] and 1997
[Magelky, 1998]. Each set of experiments consisted of(1) natural
colloid mobilization by chemical perturbations, (2) syntheticcolloid
deposition under ambient conditions,  and (3) synthetic colloid
mobilization  by chemical  perturbations. Injections of 100L of
amended groundwater were made into the uncontaminated and
contaminated zones of the aquifer using four arrays of multi-level
samplers (MLSs). Each MLS array consisted of an injection MLS
and four to six monitoring  MLSs  at down-gradient intervals of
approximately 1  m.  In 1996, three  identical  virus and colloid
injections were followed by three different chemical perturbation
injections (Table  2). In 1997, a variety of colloid  and chemical
perturbation  injections were made (Table 3).
     Sampling and Field Analysis. During both field experiments,
six depths in each of the MLSs were sampled immediately before
and  after the injections to  measure  background  and initial
concentrations (C0)  for each constituent.  After the  injections,
samples were collected daily (1996) or twice daily (1997). In the
field, sample  pH, specific conductance, dissolved oxygen, and
ferrous iron were measured.  The pH meter and electrode were
calibrated  with pH 4, 7,  and  10 buffers at  the  groundwater
temperature  and  calibrations varying by more than 5% from the
proper Nernst response were redone. The specific conductance
electrode and meter were calibrated  with  solutions  of known
conductance atthe groundwatertemperature. Calibrations varying
by more than 5% from the known conductances  were redone.
The dissolved oxygen measurements, made by Rhodazine D™
(0 to  1 mg L1) and indigo  carmine (1 to  12 mg L1) test  kits
(CHEMetrics, Inc.), were checked against solutions  purged by

-------
 Table 1.    Chemistry of the groundwater in the unconfined glacial outwash aquifer approximately 150 m downstream of the sewage
            infiltration beds at the Cape Cod site.
              constituent
unit
uncontaminated    contaminated
     zone             zone
reference
pH
specific conductance
ionic strength
temperature
dissolved oxygen
dissolved organic carbon
MBASa
Na+
K+
Mg2+
Ca2+
NH4+
Mn (dissolved)
Fe (dissolved)
cr
NO3
HCCV
SO42"
total PO43"
sediment clay content
sediment Fe(III)b
sediment phospliate
sediment foc


H-S cm"1
mM
°C
mgL'1
mgL-1
mgL-1
U.M
uM
|JM
uM
U.M
uM
uM
U.M
U.M
|JM
uM
U.M
wt%
uinol g'1
u.mol g'1


5.4 to 5.6
30 to 40
0.5
15.5
4.5 to 6.5
0.4 to 1.0
0.05
250
21
37
28
<1
0.64
0.05
230
<10
28
85
0.74
0.33±0.02
3.6±0.3
0.61±0.09
0.0001

5.8 to 6.0
250 to 330
4.0
15.0
0 to 0.5
2.0 to 4.4
0.10
1,900
200
130
210
<1
15
0.16
760
300
640
360
12
0.35±0.06
4.7±1.4
0.58 to 1.5
0.01

this study
this study
this study
this study
this study
this study
this study
this study
LeBlanc et al. [1991]
LeBlanc et al. [1991]
this study
LeBlanc et al. [1991]
LeBlanc et al. [1991]
this study
LeBlanc et al. [1991]
LeBlanc et al. [1991]
LeBlanc et al. [1991]
LeBlanc et al. [1991]
LeBlanc et al. [1991]
this study
this study
Walter etal. [1996]
Scholl and Harvey
[1992]
            a MBAS, methylene blue active substances (detergents, surfactants).
            b Sediment Fe(lll) is ferric iron extracted by Ti(lll)-citrate-EDTA-bicarbonate [Ryan and Gschwend, 1991].
nitrogen and saturated by air at the groundwater temperature.
Ferrous iron measurements, made by a 1,10-phenanthroline 0.1
to 10 mg Latest kit (CHEMetrics, Inc.), were checked against 0.1
and 1.0 mg L1 FeCI2 solutions prepared in the field laboratory.
     Injection Constituents.  Bromide, added as NaBr, was used
as a conservative tracer and  measured by ion-specific electrode.
The bromide electrode was  calibrated by 0.01, 0.1 and 1.0 mM
sodium bromide solutions before and after each  set  of
measurements. Calibration curves showing  5% variation from
the proper Nernst response were redone. Sodium hydroxide was
added after dissolution in 1 Lofdeionizedwater(MilliporeMilli-Q)
to elevate pH. Sodium dodecylbenzenesulfonate (NaDBS)
dissolved in high purity water was added to elevate the surfactant
concentration  and measured  with a  methylene blue active
substances (MBAS) test kit  (Hach Co.) with a detection limit of
0.1 mg L1.TheMBAStestkitwastestedagainstNaDBSsolutions
of known concentration, /.-ascorbic acid dissolved in deionized
water was added to elevate the reductant concentration. Ascorbic
acid concentrations were measured by UV spectrophotometry at
                  264 nm checked against standards made in uncontaminated and
                  contaminated groundwater. To reduce the ionic strength of the
                  groundwater, injections  of deionized water were made and
                  tracked by specific conductance.
                       Silica colloids (Nissan Chemical Industries, MP-1040) of
                  107±21 nm diameter were used  as the synthetic colloid during
                  the 1996 experiments. Silica colloid concentration was measured
                  by UV spectrophotometry at 340 nm (with a detection  limit of
                  about C C0"1=0.001) and turbidity (with a detection limit of about
                  C C0"1=0.01). The spectrophotometer and turbidity meter were
                  calibrated using silica colloid suspensions in the uncontaminated
                  and contaminated groundwaters.  The silica colloids were chosen
                  to  mimic  the negative surface charge found on most  natural
                  colloids and to provide a uniform colloid size distribution forwhich
                  to assess the collision efficiency.
                       To improve the detection of the colloids injected during our
                  1997 field experiment, two "tracer" colloids composed of metal
                  oxides (zirconia,  ZrO2;  titania,  TiO2)  coated  by silica were
                  developed [Ryan, Magelky, and  Elimelech, in preparation]. The

-------
                                                                                    5m
  0

  2-

  4-

  6

Im



  2-

  4-

  6

;m

  0

  2-

  4-

  6-

;m
MLS 3-11
Injection
V

5]
MLS 3-1 2
Injection
V

§


_



3A-11
1
5^
3 A- 12
5s,
|








_

1
1
4-11
1
55,
4-12




_



4A-11
5s,
s
§
4A-12
|
=51


_



5-n
51
|
5-n
§
1








_
c
Sj
1
5A-n
51
s
»A-12











(
=51
J=l


_



6
S
§
12
A
L
2
C
7
                                                                                               6-11
                                                                                                Array 1
                                                                                              ?.   Uncontaminated
                                                                                                 Zone

                                                                                              s   Contaminated
                                                                                                 Zone
                                                                                             Array 2
                                                                                               Uncontaminated
                                                                                               Zone

                                                                                               Contaminated
MLS 4- 15
Injection
_3__
^
*a
_P




4A-15

5s
5^
=!)
§




5-15

^
=5]
=a
^
^




5A-15

*%!
*5,
^
=4)
§



_
6-15
Array:
j Uncontaminated
* Zone
S Contaminated
" Zone
                   (iroundwaler How
Figure 1.    Sketch of multilevel sampler (MLS) arrays used in the field experiments at the Cape Cod site. Array numbering shown
            for 1996 experiments. For 1997 experiments, Array 1 is headed by MLS 3-11, Array 2 is headed by MLS 4-11, Array 3
            is headed by MLS 3-12, and Array 4 is headed by MLS 4-15.  Sampling ports are separated by 25 cm depths. Depths
            to the water table and the boundary  between the Uncontaminated and contaminated groundwater are shown.
zirconia particles were purchased from Aldrich Chemical Co. in
an acetic acid solution. Thetitania particles were synthesized by
the hydrolysis of titanium(IV)tetraethoxylate in a mixture of water
inethanol(1%v/v)and hydroxypropyl cellulose. The zirconia and
titania particles were coated with silica by hydrolysis of tetramethyl
orthosilicate (Table 4). The concentrations of the silica-coated
zirconia and  titania  particles were  measured by inductively-
coupled plasma-atomicemission spectrophotometryforzirconium
and titanium, providing  a much improved detection  limit of
C C0~1 < 10"5 owing to the very low background concentrations of
zirconium and titanium. A particle nebulization efficiency factor
was established by comparing  instrument response to the pure
zirconia and titania particles to acidified  solutions of Zr and Ti
ions.  The silica coatings on these particles mimic the negative
                                                    charge found on most natural particles (as did the pure silica
                                                    particles) and the  relatively  uniform size distributions foster
                                                    accurate collision efficiency calculations and modeling.
                                                         The virus used in the injections is the bacteriophage PRD1
                                                    [Olsen et al.,  1974]. The  PRD1  were radiolabeled  with
                                                    32P-phosphate[Lovelandetal., 1996]. PRD1, which was isolated
                                                    from municipal sewage, has a negative surface charge and size
                                                    typical of  many  pathogenic waterborne  viruses. Virus
                                                    concentrations were measured by liquid  scintillation counting
                                                    and checked by  standards of  known 32P activity.
                                                         Colloid Characterization. The abundance of natural colloids
                                                    was measured by turbidity. Turbidity was converted to a colloid
                                                    concentration by measuring the mass of colloids in samples of
                                                    known turbidity.  The natural colloids were also characterized by

-------
Table 2.
Table 3.
Summer 1996 injection schedule. Constituents of natural colloid mobilization and silica colloid and bacteriophage
PRD1 deposition and recovery experiments. C0 is the concentration measured in samples withdrawn immediately
after injection.                                                     ,   .   .  ,              ,   .  ,  ,
      J                                                      uncontarmnated          contaminated
                                            injectate               C0                    C0
     injection     array    constituents        concentration          6.4 m depth            8.7m depth
natural colloid
mobilization


synthetic colloid
deposition


synthetic colloid
recovery


1
2
•i
j
1
2
3
1
2
3
NaOH
NaBr
NaDBS3
NaCl
ascorbic acid
NaBr
PRD1
SiO2 conoids
NaBr
PRD1
SiO2 conoids
NaBr
PRD1
SiO2 conoids
NaBr
NaOH
NaBr
NaDBS
NaBr
ascorbic acid
NaBr
pH12.5
l.SOmM
0.57mM
2.56mM
1.82mM
l.SOmM
2.6xl06cpmL-1
500 mg L-1
l.SOmM
2.1xl06cpmL-1
500 mg L-1
l.SOmM
2.2xl06cpmL-1
500 mg L-1
l.SOmM
pH12.5
l.SOmM
57 mM
l.SOmM
1.82mM
l.SOmM
pH12.3
1.18mM
0.34mM
2. 13 mM
0.63mM
1.33mM
390 mgL-1
0.79mM
400 mg L-1
l.OSmM
LSxlO'^cpmL1
210 mgL-1
0.90mM
pH11.7
0.77mM
52 mM
1.36mM
1.76mM
0.86mM
pH11.9
1.46mM
0.66 mM
2.47 mM
1.73 mM
1.56mM
1.9xl06cpmL-1
470 mg L-1
0.90mM
l.SxlO^cpmL-1
480 mg L-1
1.12mM
310 mgL-1
1.02 mM
pHll.8
1.04mM
52 mM
1.07 mM
2.09 mM
l.llmM
3 NaDBS is sodium dodecylbenzene sulfonate, an anionic surfactant.
Summer 1997 injection schedule. Constituents of natural colloid mobilization and silica-coated zirconia and titania
colloid deposition and recovery experiments. C0 is the concentration measured in samples withdrawn immediately
after injection. Silica-coated zirconia and titania colloids are shown as Si/ZrO2 and Si/TiO2, respectively.
uncontaminated contaminated
injectate Co C0
injection array constituents concentration3 6.4m depth 8.7m depth
natural colloid
mobilization


synthetic colloid
deposition



1st synthetic
colloid recovery

2IKl synthetic
colloid recovery
1
3
4
1
2
3
4
I
3
3
NaOH
NaBr
deionized water11
NaDBS'
NaBr
Si/ZrO,
NaBr
Si/ZrO2
NaBr
Si/ZrO,
NaBr
Si/ZrO2
SL'TiO,
NaDBS
NaBr
NaOH
NaBr
deionized water
NaOH
NaBr
pHll
l.OOmM

29 mM
l.OOmM
24 9/230 ppm Zr
l.OOmM
26 1/275 ppm Zr
1.00 mM
246/234 ppm Zr
1.00 mM
61/124 ppm Zr
93/140 ppm Ti
72 uM
l.OOmM
pHll
1.00 mM

pHIO
1.00 mM
pH 10.3
0.75 mM

43 mM
0.34 mM
190 ppm
l.llmM
191 ppm
0.99 mM
283 ppm
0.92 mM
62.7 ppm
41.0 ppm
72 uM
1.04 mM
pH 10.8
0.87mM

pH9.7
1.09 mM
pIilO.8
0.72 mM

llmM
0.75 mM
51. 9 ppm
l.llmM
159 ppm
0.69 mM
240 ppm
0.94 mM
33.0ppm
41.1 ppm
72 uM
1.11 mM
pH 1 1 .2
0.81 mM

pH9.7
l.OlmM
                " Injectate concentration in uncontaminated and contaminated zones shown as "uncont/cont" where the concentrations were
                different in the two injectates.
                " Millipore Milli-Cf-'-prepared water used as deionized water
                c NaDBS is sodium dodecylbenzene sulfonate, an anionic surfactant.

-------
Table 4.     Size of titania and zirconia particles before and after coating with silica. Size measured by dynamic light scattering
            reported as the mean diameter ± one standard deviation for a Gaussian distribution of particle sizes.

particle
zirconia
titania
uncoated
diameter
(nm)
1 10 ±34
1030±140
silica-coated
diameter
(nm)
130 ±49
1080±160
silica coat ing
thickness3
(nm)
10
50
               3 Coating thickness estimated as one-half the difference between the
               silica-coated and uncoated diameters.
SEM  and energy-dispersive  x-ray spectroscopy (EDX) after
being trapped on filters, mounted on aluminum stubs, and gold-
or carbon-coated. The electrophoretic mobility of the colloids and
virus  were determined by microelectrophoresis (Brookhaven,
model  ZetaPlus)  in  uncontaminated  and  contaminated
groundwater and  groundwaters  amended by  the chemical
perturbations. The consistency of microelectrophoresis
measurements was checked by measuring the zeta potential of
polystyrene latex  microspheres. During  the elevated  pH
mobilization experiments in 1997, dissolved organic carbon was
measured in 0.45|um-filtered samples by combustion on a platinum
catalyst (Shimadzu TOC-5000). The  organic carbon  analyzer
response  was calibrated with potassium biphthalate solutions.
Total  organic  carbon concentration  was also  measured  by
absorbance of light at 254 nm in unfiltered samples.
     Aquifer  Sediments.  Cores were  obtained   from the
uncontaminated  and contaminated zones of the aquifer about
5 m east of the injection wells used in the field experiments. The
sediments were impregnated with epoxy resin and thin-sectioned
for  electron microprobe (JEOL, model 8600,  15 kV accelerating
voltage) and SEM (backscatter electron detection) analysis. The
area of surfaces covered by ferric oxyhydroxide coatings were
estimated by observing the fraction of surface coatings for 400
grains in the uncontaminated and contaminated zone sediments.
The mineralogy  of the fine  particles of the  sediments was
determined by powder x-ray diffraction (XRD). Total Fe, Al, and
Mn were  measured by digestion in concentrated  HF/HNO3
solutions.  Free Fe, Al, Mn were measured by Ti(lll)-citrate-EDTA-
bicarbonate extraction [Ryan  and Gschwend,  1991]. The total
digestions and  free element extractions were checked  by
measuring the Fe,  Al, and Mn content of pure oxides of these
elements. The streaming potentials of  aquifer grains  were
measured in the laboratory of Dr. Philip R. Johnson at Notre
Dame University.
     Data Quality Assurance. The data collected in the field were
subject  to careful quality assurance procedures outlined in the
Quality Assurance Project Plan for  Colloid Mobilization and
Transport in Contaminant Plumes, a report prepared by the
project investigators at the beginning of the project.   All of the
data presented in  this  report met the data  quality  assurance
criteria.
     Specific quality assurance procedures have been included
with the details of  analysis above. In  addition to these specific
procedures, many  general quality assurance procedures were
followed. Analytical methods used to characterize the sediment
and groundwater  samples  were tested  using  standards and
spiked samples. Selected samples were measured on alternate
instruments available in nearby laboratories  (particle size, zeta
potential)  for analyses that cannot be absolutely calibrated or
tested foraccuracy. The influence of contamination from reagents
and laboratory environment was assessed using reagent blanks
and method blanks. Instrumentvariability was tested with internal
standards. Measurements were checked by mass balance. During
fieldwork, field blanks were  included in the samples to assess
contamination of samples  by sample containers, sampling
handling, storage, and shipping. Preliminary studies were carried
out to  determine the primary sources of error in analytical
procedures. In both the laboratory and the field, replicate samples
were tested at various times to assess the effect of sample aging
and storage on results.
     Calculation of Attenuation, Collision Efficiency, Recovery,
and Interaction Forces. The relative breakthrough (RB, %) of an
injected constituent is a ratio of the time-integrated mass of the
constituent relative to that of the  conservative  tracer. Collision
efficiencies (a) were  calculated with longitudinal dispersion  for
pulse inputs of PRD1 and silica colloids [Harvey and Garabedian,
1991], In the collision efficiency calculations, the single collector
efficiencies [Yaoetal., 1971; Rajagopalan and Tien, 1976] were
calculated using only the convective-diffusion and sedimentation
contributions. An average grain diameter of 0.6 mm, a porosity of
0.39, and fluid velocities of 0.7 m d~1 forthe uncontaminated zone
and 0.4 m d*1 for the contaminated zone were used in  these
calculations. Fractional recoveries of virus and silica colloids
following the chemical perturbations were estimated as  the
quantity of virus or colloid recovered during the second injection
divided by the quantity of virus or colloid immobilized during the
first injection [Pieper et al., 1997].
Results
     Colloid Characterization.  The mobilized  natural colloids
ranged in size from <0.1 to 10|am, were platy in shape, and were
composed  primarily  of  silica  and  aluminum  with traces of
phosphorus and  iron. No significant differences were observed
for colloids mobilized in the uncontaminated  or contaminated
zones orfor colloids mobilized by different chemical perturbations.
Mineral identification by XRD revealed quartz, feldspar, kaolinite,
illite/muscovite, and  smectite, but no crystalline  ferric
oxyhydroxides. Microelectrophoresis analysis revealed a pHpzc
of approximately 2.2±0.3. Above  pH 9.5,  the zeta potentials of
the colloids reached a plateau of about -45 mV.
     The silica and silica-coated colloid surfaces were negatively
charged from about pH 3 to the ambient pH of the uncontaminated
groundwater  (Tables 5  and 6).  The silica  and silica-coated
colloids were slightly more negatively charged in the contaminated
groundwater than in the uncontaminated groundwater.
     The PRD1 surface was negatively charged from pH 3.2 to
the ambient pH of the uncontaminated groundwater (Table  7).
PRD1 was slightly more negatively charged in the contaminated
groundwaterthan in the uncontaminated groundwater. The PRD1

-------
Table 5.
Silica-coated zirconia colloid zeta potentials under ambient and perturbed conditions.

sample
silica-coated ZrO2 colloids
uncontam groundwater
LSmMNaBr






silica-coated ZrO2 colloids
uncontam groundwater





silica-coated ZrO2 colloids
con tarn groundwater
LSmMNaBr






silica-coated ZrO2 colloids
con tarn groundwater






conditions
pH 3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
11.0
ambient
1 mM NaBr
72 uM NaDBS
49 mM NaDBS
1 mM AscAc
pHlO.O
pH 1 1.0
pH 3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
11.0
ambient
1 mM NaBr
72 uM NaDBS
49 mM NaDBS
1 mM AscAc
pHlO.O
pH 1 1.0
zeta potential
(mV)
-16±4
-18±3
-19±4
-21±2
-20±2
-25±7
-21±2
-23±2
-35±5
-21±2
-23±1
-22±2
-40±2
-24±3
-23±2
-35±4
-20±4
-20±2
-21±2
-19±1
-22±2
-23±2
-21±3
-24±2
-26±2
-20±1
-24±2
-22±1
-40±2
-25±4
-24±1
-26±2
Table 6.     Silica-coated titania zeta potentials under ambient and perturbed conditions.
                                                                         zeta potential
              sample	conditions	(mV)
titania (TiO2) colloids
l.SmMNaBr




silica-coated TiO2 colloids
l.SmMNaBr





pH 3.2
4.0
4.5
6.1
9.6
10.0
pH 3.1
3.8
4.1
5.9
7.6
9.5
11.2
9±2
4±3
-8±1
-2 3 ±2
-28±5
-26±2
1+2
-7±3
-13+2
-21+2
-22±4
-27±4
-25±2

-------
Table 7.     Bacteriophage PRD1 zeta potentials under ambient and perturbed conditions.
sample
PRD1
uncontam groundwater
l.SmMNaBr
conditions
pH 3.2
4.1
5.2
6.1
zeta potential
(mV)
-8±3
-11±3
-17±2
-25±4
             PRD1
             contam groundwater
             l.SmMNaBr
pH
6.4
-27±4
zeta potentials are slightly more negative than those measured
for PRD1 in a calcium phosphate solution [Bales et al., 1991].
     Sediment Characterization.   XRD  revealed that  the
sediments were composed  mostly of quartz,  feldspars,  and
traces of clay minerals (kaolinite, illite/muscovite, smectite)  and
iron  oxyhydroxides.  Electron microprobe analysis of sediment
thin sections from both the uncontaminated and contaminated
zones  revealed scattered coatings on the  quartz grains  that
contained iron  oxyhydroxide and clay minerals  in both  the
uncontaminated and contaminated zones (Figure 2). The coating
coverage was estimated at 3.0±10.0% of the uncontaminated
grain surfaces and 3.5±11.1 % of the contaminated grain surfaces.
Most of the grains were uncoated and a  small fraction of the
grains were coated an average of about 50%. This small amount
of surface coverage agrees well with the relatively low sediment
Fe(lll) concentrations measured by reductive extraction in the
aquifer sediments (Table 1). The zeta potential of contaminated
sediments in the contaminated groundwater is significantly more
negative than that  of the uncontaminated sediments in  the
uncontaminated groundwater (Table 8).   The  natural colloids
mobilized fromthe uncontaminated sediments were characterized
by a pH^ of about 2.3 and the chemical perturbations increased
their negative charge (Table 9).
     Natural  Colloid Mobilization by Chemical Perturbations.
Over the two field  seasons, two  injections of elevated  pH
groundwater were performed at approximate injection pH values
of 11 (1997) and 12.5 (1996). The amount of colloids mobilized
increased with increasing pH (Table 10). Colloid mobilization in
the uncontaminated zone always exceeded colloid mobilization
in the contaminated zone. The pH increase was buffered to a
much greater extent in the contaminated zone. The appearance
of the elevated pH plume and the mobilized colloids  lagged the
tracer breakthrough in these experiments. Details of the effect of
pH on natural colloid mobilization will be presented in Ryan, Ard,
Magekly, and Elimelech [in preparation].
     During the  1996 experiment, the lower concentration of
NaDBS (0.57 mM, 200 mg L1),  about half the  critical micelle
concentration, caused greater mobilization of natural colloids in
the uncontaminated zone. Owing to the presence of bubbles at
the high NaDBS concentration (50 mM, 1 %), we were not able to
accurately measure turbidity during the 1997 experiment. At the
lower concentration, a measurable amount of the  NaDBS was
attenuated, presumably by sorption to the aquifer sediments, but
the overall transport of the NaDBS plume was not significantly
retarded; thus, colloid mobilization coincided or slightly lagged
the passage of the tracer and NaDBS pulses.
     During the 1996 experiment, an increase  in the ascorbic
acid concentration produced much greater colloid mobilization in
the uncontaminated zonethan in the contaminated zone. Ascorbic
               acid was  attenuated to C C0~1 levels of 0.05 and 0.5 in the
               uncontaminated  and  contaminated zones of the  aquifer,
               respectively, but no significant retardation was  observed. The
               mobilized colloids appeared at the same time or slightly behind
               the tracer and ascorbic acid pulses. Iron(ll) concentrations well
               above the field detection limit of 1.8 |aM were detected only in the
               contaminated zone.
                    During the 1997 experiment, a decrease in the ionic strength
               of the groundwater produced a small amount of colloid mobilization
               with slightly greater mobilization occurring in the uncontaminated
               zone. The deionized water "plume" moved through the array at
               the same rate as the bromide tracer moved through the array in
               later experiments and the mobilized colloids always appeared at
               the same time  or slightly after  the depression of specific
               conductance caused by the deionized water  pulse.
                    Synthetic Colloid  and Virus Deposition. During  the  1996
               experiments, silica colloids and bacteriophage PRD1 displayed
               measurable breakthroughs over the first meter  of  transport
               (Figure 3). Clear breakthrough curves were not detected further
               down-gradient  [Ard, 1997]. The silica colloids and PRD1  were
               less attenuated  in the  contaminated  zone than  in  the
               uncontaminated zone (Table 11). Collision efficiencies calculated
               with and without dispersion varied by up to a factor of 3.4 with the
               collision efficiency  calculated with  dispersion always the larger
               value. Details of these experiments are presented in Ryan, et al.,
               [1998].
                    During the  1997 experiments, the  transport  of the silica-
               coated zirconia displayed similar trends in the  same arrays -
               relative  breakthroughs  were  higher and collision efficiencies
               were lower in the contaminated zone of the aquifer. Again,  clear
               breakthroughs were not evident beyondthel mtransportdistance
               [Magelky, 1998]. In a special experiment conducted  in 1997, both
               the silica-coated zirconia and titania particles were injected into
               array 4 following the injection of 29 mM (1%) NaDBS to mobilize
               natural colloids.  Under  these  conditions, transport through the
               uncontaminated and contaminated zones was similar. The silica-
               coated zirconia colloids displayed collision efficiencies that  were
               substantially lowerthan  those of the silica-coated titania colloids.
                    Synthetic Colloid  and Virus Mobilization. The amount of
               colloids and viruses mobilized by pH  elevation increased as the
               pH of the  injection was increased from  10 to 12.5 (Table 12).
               Generally, the pH increase was  more  effective at  mobilizing
               colloids in the uncontaminated zone (Figure 4). The migration of
               the NaOH plume lagged significantly behind the tracer, but during
               the pH  11  injection in  1997,  some of the  mobilized colloids
               appeared slightly ahead of the pH increase at the 1 m transport
               distance (Figure 5). Dissolved organic carbon mobilized by this
               pH increase also appeared slightly ahead of the pH  increase.
               Details of this work will be provided in Ryan, et al. [in  preparation].

-------
                                  Core  13
                                                                                 (a)
                                                     FeOOH
                                0001    15.8KU
X55     100HID
                                                                                 (b)
                                 300 -

                                 250

                               en 200
                              •£3

                               8 15°
                                 100 -
                                                                                 (c)
                                     02468

                                                   energy (keV)

Figure 2.    Electron microprobe images of resin-impregnated thin section of contaminated zone sediments of the Cape Cod glacial
            outwash aquifer showing AI-, Si- and Fe-rich coating on quartz grain at two different scales: (a) scale bar, 100|am;
            magnification, 55 times; (b) scale bar, 10 jam, magnification 500 times. Dark grains are quartz; bright rims contain Al, Si,
            and Fe. EDX scan of coating (c).

-------
Table 8.
Table 9.
Aquifer sediment zeta potentials under ambient and perturbed conditions.
                                                         zeta potential
sample
uncontam sediment in
uncontam groundwater




contam sediment in
contam groundwater




conditions
ambient, pH 5. 8
1 mM NaBr
72 uM NaDBS
49 mM NaDBS
pHlO.O
pHll.O
ambient, pH 6. 2
1 mM NaBr
72 uM NaDBS
49 mM NaDBS
pHlO.O
pH 1 1.0
(mV)
-22±1
-23±1
-3 Oil
-43±1
-3 Oil
-31±2
-27±1
-27±1
-24±1
-40±1
-31±1
-35±1
Natural colloid zeta potentials under ambient and perturbed conditions.

sample
natural colloids mobilized
from uncontam sediments
suspended in deionized water







natural colloids mobilized
from uncontam sediments
pH5.8



conditions
pH 2.3
2.9
3.7
5.6
6.2
7.8
8.8
10.1
11.4
12.0
NaDBS 0
0.57 mM
2.9 mM
14mM
29 mM
zeta potential
(mV)
0±8
-4±2
-18±5
-26±3
-33±4
-37±3
-44±2
-43±4
-42±7
-45±8
-28±6
-27±2
-33±5
-45±6
-46±4
     During the 1996 experiment, NaDBS was more effective at
mobilizing colloids and viruses in the contaminated zone than in
the uncontaminated zone. A comparable experiment was not
carried out during the 1997 field  season. SEM examination of
filtered particles revealed at least 90% silica colloids; therefore,
no correction for mobilization of natural colloids was applied. The
migration of the surfactant closely matched the tracer migration.
The  mobilized colloids  and viruses  appeared at the first MLS
concurrently or slightly behind the surfactant peak concentration.
     During the 1996  experiment, ascorbic acid  was  more
effective  at mobilizing  the  silica colloids than the viruses. A
comparable experiment was not conducted during the 1997 field
season. SEM examination of filtered particles revealed at least
90% silica  colloids; therefore, no correction for mobilization of
natural colloids was applied. The transport of ascorbic acid was
attenuated, especially  in the  uncontaminated zone,  but no
significant retardation of ascorbic acid occurred. The mobilized
colloids and viruses appeared at about the same time or slightly
                                                 after the appearance of the ascorbic acid peaks at the first MLS.
                                                 The ascorbic acid injection produced a significant increase in the
                                                 Fe(ll) concentration in the contaminated zone, but no significant
                                                 increase above the detection limit in the uncontaminated zone.
                                                 Dissolved oxygen concentrations were slightly depressed in the
                                                 uncontaminated zone during the ascorbic acid injection, but not
                                                 significantly different in the contaminated zone.
                                                      During the 1997 experiments, a decrease in ionic strength
                                                 brought about by an injection of 100 L of deionized water mobilized
                                                 more colloids  in the uncontaminated zone of the aquifer. The
                                                 mobilized colloids appeared at the 1 m transport distance at the
                                                 same time or slightly behind the deionized water pulse.

                                                 Discussion
                                                      Silica and Silica-Coated Colloid, PRD1, and Aquifer Grain
                                                 Zeta Potentials During Injections. The negative zeta potentials of
                                                 the silica and silica-coated colloids reflect the predominance of
                                                 deprotonated surface hydroxyls at pH values above  the pH
                                                          10

-------
Table 10.   Amount of natural colloids mobilized by chemical perturbation injections.
                                                                   mass of colloids mobilized3
                  experiment          conditions    aquifer zone               (mg)
11
elevated pH
(NaOH) 12.5
0.57mM
^vat^d surfactant
NaDBS 29mM
elevated reductant !-° lllM
AscAc
decreased 
-------
Table 11.    Summary of relative breakthroughs (RB) and collision efficiencies (a) calculated with and without dispersion for silica
             and silica-coated colloids and PRD1.
colloid
PRD1
silica
silica-coated ZrO2
silica-coated ZrO2 after 1%
NaDBS
silica-coated TiO2 after 1%
NaDBS
aquifer
zone
uncontam
contain
uncontam
contam
uncontam
contam
uncontam
contam
uncontam
contam
replicate
arrays
3
3
3
3
3
3
1
1
1
1
RB
(%)
2.5±1.7
4.3±1.4
13±4
37±16
1.3±1.5
34±42
78
76
11
12
a
with dispersion
0.032±0.016
0.016±0.005
0.023±0.009
0.0056±0.0034
0.039±0.013
0.012±0.008
0.0021
0.0017
0.0077
0.0090
Table 12.    Recovery of silica and silica-coated colloids and PRD1 over the first meter of transport. Two pairs of recoveries are
             listed for the silica colloids at an elevated pH of 12.5 - the first pair is for silica colloids measured by UV absorbance
             and the second pair is for silica colloids measured as  0.6 times the sample turbidity to account for natural colloids.
chemical
perturbation conditions
elevated pH 1 0.0
NaOH
11.0
12.5





elevated 0. 57 mM
surfactant
NaDBS


elevated 1 .0 mM
reductant
AscAc


decreased <5 u.S cm1
ionic strength
deionized water

colloid
silica-coated ZrO2
silica-coated ZrO2

PRD1

silica (UV)

silica (0.6turb)

PRD1


silica (UV)

PRD1


silica (UV)

silica-coated ZrO2

aquifer zone
uncontam
contam
uncontam
contam
uncontam
contam
uncontam
contam
uncontam
contam
uncontam
contam

uncontam
contam
uncontam
contam

uncontam
contam
uncontam
contam
recovery
(%) '
0.00
0.10
1.5
0.70
100
78
240
120
103
37
0.5
49

3.5
15
3.8
24

51
27
0.50
0.10
                                                           12

-------
       z
       o
       N
       Q
       W
       H
       <
       Z
       H
       z
       o
       u
       z
     400 -
     200 -
        0
     200 -
     100 -
                                                                 PRD1
           HH
           HH
             O
           p
           b
     -o— silica (UV)
     —B— silica (0.6 turb)
-ffl—rn--m_
       Z
       O
       N
       Q
       M
       H
       <
       Z
       H
       Z
       O
       u
Figure 4.
           J  400 -
            ^ 200 H
           &     0
           HH
           HH
            Ol
                                                       PRD1
                                   ^=*
                                             - silica (UV)
                                               silica (0.6 turb)
                                               »""        •--•
                                                	^mm.S'Sm~~Sm
       10 -I
        8 -
        6 -
                                                                    PH
 AAAAAAAA
                                                     bromide
                      0
                        4     6    8    10    12    14    16
                                time (d)
Mobilization of silica-coated zirconia colloids after deposition by elevation of pH (pH 11) at 1 m transport distance during
1997 field experiments. Silica-coated zirconia colloids measured as Zr  by ICP-AES. Bromide tracer breakthrough
presented as normalized concentration.
                                               13

-------
         E
          OH
         P
  8
  7
  6
  5
 40
 30
 20
 10
  0
         J2 4000
          OH
         3 2000
          JH
         I
         O
         o
         Q
2.0 -I
1.5 -
1.0 -
0.5 -
0.0
          1 o   1.0  -
          O
          O    0.5  -
          '&    0.0  -
                       0
                         6    8    10   12    14    16    18
                             time (d)
Figure 5.   Mobilization of silica-coated zirconia particles by an increase in pH in the uncontaminated zone during the 1997 field
          experiments. The breakthrough of zirconium, turbidity, and dissolved organic carbon (DOC) slightly precede the advance
          of the pH increase.
                                                 14

-------
value of about 2 to 2.5 reported for silica [Parks, 1965].  The
similarity between the pHpzc and the zeta potentials of the silica
and silica-coated colloids indicate thatthe silica coating thickness
of 10 to 60 nm was sufficient to mask  the underlying  surface
properties of zirconia and titania.
     The pHpzc for PRD1 in the uncontaminated groundwater is
less than 3.2, similarto a pH   value of less than 3.9 in a  calcium
phosphate solution. On the basis of studies by Penrod et al.
[1996] relating the  pHpzc  values of bacteriophage  MS2  and
lambda to the composition of their protein capsids, these low
pHpzc values for PRD1 indicate that deprotonatedcarboxyl groups
in amino acids dominate the  surface speciation of the PRD1
protein capsid.
     In the contaminated groundwater, the silica and silica-
coated colloids and PRD1 were slightly more negative than in the
uncontaminated groundwater. For PRD1, the change  in zeta
potential may  simply  be caused  by  the  higher pH of the
contaminated groundwater, although changes in virus  surface
charge may also be attributed to surfactants [Small and Moore,
1987]  and fulvic acid  [Bixby and O'Brien, 1979]. The silica
colloids,  however,  appear to have  reached a constant zeta
potential at about pH 4.4,  so the more  negative zeta potential
measured in the contaminated groundwater cannot be attributed
to the  higher pH of the contaminated groundwater.  The zeta
potential of the silica  colloids  must have  been made more
negative by adsorption of anions;  e.g., organic  matter  and
phosphate. While extensive adsorption of organic matter to silica
at pH near 6.0 is unlikely owing to electrostatic repulsion [Davis,
1982], calcium and magnesium in the contaminated groundwater
may enhance organic matter adsorption by reducing the negative
charge of the organic matter.
     The measuredzeta potentials ofthe aquifergrains represents
the net charge of  a heterogeneous surface made  up of the
underlying quartz grains,  ferric oxyhydroxide and clay  mineral
coatings, and adsorbed organic matter  and phosphate. At the
ambient pH of the  groundwater, quartz is negatively charged,
ferric oxyhydroxide is positively charged, and clay minerals are
negatively charged overall with positively charged edges [Parks,
1967]. Adsorption of organic matter and phosphate to the positively
charged surfaces reverses the surface charge [Liang and Morgan,
1990]. The net negative surface potential ofthe uncontaminated
sediment suggests that the ferric oxyhydroxide coating on the
quartz grains  is patchy, as suggested by thin sections of Cape
Cod  aquifer grains examined  by Coston et al. [1995] and this
study (Figure 2). We attribute the more negative zeta potential of
the contaminated sediment to the much higher organic matter
content ofthe contaminated sediments (Table 1). The phosphate
content ofthe contaminated sediments is also elevated,  but only
by a  maximum factor of about 2.5.
     Silica and Silica-Coated Colloid  and PRD1 Deposition
Behavior.  Most studies of virus attachment to mineral grains
concludethatelectrostaticforces dominate the interaction between
virus and grain surfaces [Gerba,  1984; Murray and Parks, 1980;
Loveland etal., 1996; Penrod etal., 1996; Redman etal., 1997],
If electrostatic forces dominated colloid  and virus deposition in
these experiments,  the zeta  potential  data should  provide a
qualitative explanation for the  observed  deposition behavior.
     Both the silica and silica-coated colloids and viruses were
transported through the contaminated zone more  readily than
through the uncontaminated zone. The  greater abundance  of
organic matter in the contaminated sediments appears  to have
masked the ferric oxyhydroxide coatings,  giving the contaminated
sediments  a greater  negative  zeta potential than the
uncontaminated sediments. Consequently, when the negatively
charged  colloids  and viruses  interact with the contaminated
sediments, they experience greater repulsion, resulting in collision
efficiencies lower than those measured in the uncontaminated
zone.
     The zeta potential data and the ferric oxyhydroxide coatings
detected in the thin sections indicate that ferric oxyhydroxides in
the Cape Cod sediments enhance colloid and virus attachment
and organic matter in the sewage plume inhibits colloid and virus
attachment. The presence of positively charged oxides limits the
transport of viruses and bacteria because these "biocolloids" are
typically negatively charged at the ambient pH of most waters and
readily attach to positively charged surfaces [Murray and Parks,
1980; Moore etal., 1981; Farrah and Preston, 1991; Scholl and
Harvey,  1992; Mills et al., 1994;  Loveland  et al., 1996]. For
mineral colloids, Johnson et al. [1996] showed that the transport
of silica colloids in a porous media consisting of mixtures of clean
and  iron oxide-coated quartz  sand depends strongly on the
amount of iron oxide coating.  Organic matter of natural and
anthropogenic (e.g., sewage, surfactants) origin hinders virus
attachment to mineral surfaces [Charney etal., 1962; Burge and
Enkiri,  1978; Sobseyetal., 1980; Gerba etal., 1981; Moore etal.,
1981 ;1982; Atherton and Bell, 1983; Fuhsetal.,1985], presumably
by adsorbing to and masking virus attachment sites. Similarly,
the transport of ferric oxide colloids is  enhanced  by natural
organic  matter in quartz sands [Amirbahman and Olson 1993;
Kretzschmar et al., 1995].
     The differences between silica and silica-coated colloid and
virus deposition are more subtle and not statistically significant,
but a qualitative explanation of their deposition  behavior based
on the presence of ferric oxyhydroxide coatings and measured
zeta potentials fits the trend of the mean collision efficiencies
measured in the experiments.  In  both groundwaters, the silica
and silica-coated colloids possess slightly greater negative zeta
potentials, resulting in lower overall collision efficiencies.  The
close dependence ofthe measured collision efficiencies on the
zeta potentials ofthe PRD1, silica  colloids, and aquifer grains
reinforces the conclusion that electrostatic forces dominate the
transport behavior of viruses in porous media.
     In one  1997 experiment, the transport of silica-coated
zirconia and titania colloids was  measured in  a simultaneous
injection with dodecylbenzene sulfonate. The collision efficiency
calculated for the silica-coated titania particles was about 4-5
times greater than that calculated for the silica-coated zirconia
particles  (Table 11), indicating greater repulsion between colloid
and  grain for the silica-coated  titania  particles. This result
qualitatively agrees with the zeta potentials measured for the two
colloids - the silica-coated titania colloids have greater negative
zeta potentials than the  silica-coated  zirconia colloids near the
ambient pH ofthe groundwater (Tables 5,6).
     Grain Surface Heterogeneity and  Collision Efficiencies.
The  aquifer grain  surfaces are primarily made up of patchy
coatings  of ferric oxyhydroxides on the underlying quartz grains.
At the ambient pH ofthe groundwaters, the negatively charged
PRD1 should be collected by the ferric oxyhydroxide coating and
repelled  by the exposed quartz  [Loveland et  al., 1996].  The
measured  collision efficiencies  should reflect the summed
contribution of ferric oxyhydroxide and quartz  surfaces [Song et
al., 1994; Johnson etal., 1996]. Using PRD1 collision efficiencies
oiaFe0x =1 forferric oxyhydroxide (owing to electrostatic attraction)
and  a^z=6x10'3 for quartz surfaces [Bales et  al., 1991], we can
                                                          15

-------
estimate the fraction of the surface area coated  by ferric

oxyhydroxide, fFe0x, using the following equation:
         ^measured   JFeOx ^FeOx ~"~ Jqtz
(1)
This  equation gives  estimates of fFeOx = 0.026  for  the

uncontaminated zone and fFe0x = 0.010 for the contaminated

zone. The  uncontaminated zone  fFe0x estimate is  in close
agreement with the 3.0% surface coverage detected by electron
microprobe in the thin sections, butthe contaminated zone fFe0x
estimate is  lower than the 3.5% surface  coverage detected. No
significant  difference  between the  surface coverage in  the
uncontaminated and contaminated zones was  observed in the
thin  sections;  therefore, the lower contaminated zone  fFe0x
estimate is attributable to the masking  of ferric oxyhydroxide
coatings by organic matter and phosphate adsorption.
     Colloid and Virus Mobilization by Elevated pH. The NaOH
injection was  designed to reverse  the charge on the ferric
oxyhydroxide coatings  by  raising pH above the pHpzc of ferric
oxyhydroxides [Parks,  1967]. The amount of colloid and PRD1
mobilization increased with increasing pH. Although the precision
of the silica colloid recovery measurement  is  clouded by the
presence of the natural  colloids, it appears that similar amounts
of PRD1 and  silica colloids were released and that release
occurred more readily in the uncontaminated zone.
     In  the uncontaminated zone, the lack of buffering  in the
groundwater and sediments resulted  in a greater increase in pH
and  greater release of PRD1  and silica  colloids than in  the
contaminated zone. The pH in the uncontaminated zone peaked
at nearly 10 at the 1-m distance (injection pH 11.7), while the pH
in the contaminated zone peaked at only 8.5 at the 1-m distance
(injection pH 11.5) (Figure 4). An increase of pH to 10 is sufficient
to reverse the surface charge of any ferric oxyhydroxide, but an
increase to pH 8.5 may not exceed the pHpzc of some ferric
oxyhydroxides. The greater release of colloids and viruses in the
uncontaminated zone must be attributed to the increase in pH
well  in excess  of the pHpzc of the ferric oxyhydroxide coatings.
     In the contaminated zone of the Cape Cod  aquifer, Bales et
al. [1995] found that injection of a pH 8.3 solution containing an
unspecified concentration of phosphate  for buffering effectively
re-mobilized PRD1 (a  fractional recovery was not measured).
This injection caused the detachment of PRD1 even when the pH
increase down-gradient of the injection was only slightly above
the ambient pH. It is likely that the phosphate augmented the
charge reversal caused by the pH increase by adsorbing to the
ferric oxyhydroxide coatings.
     Colloid and Virus Mobilization by Surfactant Addition. NaDBS
was  added  to alter the  ferric oxyhydroxide surface charge and
promote release.  NaDBS  is an anionic surfactant that readily
adsorbs to positively charged oxide surfaces and reverses surface
charge  by  hemimicelle formation  [Dick et al., 1971]. Similar
surfactants  have been shown to mobilize natural colloids [Ryan
and Gschwend, 1994] and cause permeability reduction through
colloid mobilization [Allred  and Brown, 1994]. The high NaDBS
injection concentration  made it difficult  to determine with any
precision the amount of NaDBS  lost to aquifer sediments by
adsorption.  Only a small fraction of the NaDBS injected would be
required to saturate the ferric oxyhydroxide surfaces with adsorbed
NaDBS  [Pieper et al., 1997]; however,  NaDBS was much less
effective at mobilizing PRD1 and silica colloids than the increase
in pH. Similarly, Bales et al. [1991 ] showed that 1 % Tween 80 and
2.5% beef extract solutions were only  marginally effective at
mobilizing PRD1 and MS2, another bacteriophage, relative to an
increase in pH to 8 in a sodium phosphate solution. Bales et al.
[1991] speculated that the  high ionic strength of the surfactant
and beef extract solutions inhibited virus detachment. Ouraddition
of NaDBS to the Cape Cod  groundwater caused an increase in
ionic strength (0.57 mM) less  than that  caused by the sodium
bromide tracer (about 1.5 mM).
     NaDBS was much more  effective at mobilizing PRD1 and
silica colloids in the contaminated zone. In a previous experiment,
Pieper et al. [1997] similarly observed that a 25 mg L1 mixture of
linear alkylbenzenesulfonat.es (LAS) recovered 87% of the injected
PRD1 in the contaminated zone and only 2% in the uncontaminated
zone. The abundance of organic matter in the contaminated zone
must reduce the amount of surfactant  needed to  reverse the
charge of ferric oxyhydroxide  surfaces.  In this experiment, the
higher concentration of NaDBS (0.57 mM; 200 mg L1) did not
improve recovery.  The results suggest that  PRD1 and silica
colloids are more strongly bound in the uncontaminated zone.
     Colloid and Virus Mobilization by  Reductant  Addition.
Ascorbic  acid was  added  to  remove the ferric oxyhydroxide
coatings  by  reductive dissolution, resulting  in  the release of
PRD1 and silica colloids attached to the coatings. Ryan and
Gschwend [1994] observed that reductive dissolution by ascorbate
mobilized  natural colloids  from a similar ferric oxyhydroxide-
coated quartz sand as long  as increases  in  the ascorbate
concentration did  not raise  ionic strength to a level too high to
inhibit release. Inthisexperiment, ascorbicacidaddition promoted
PRD1 and silica colloid release that was  somewhat comparable
to the surfactant  addition, but less than  that  caused by the  pH
increase. The amount of ascorbic acid injected in this experiment
was similar to the amount promoting the maximum colloid release
in the experiments of Ryan and Gschwend [1994].
     The  amount of PRD1 and  silica  colloid  release varied
inconsistently in these experiments. Ascorbic acid appeared to
be effectively dissolving ferric oxyhydroxides in the contaminated
zone because the Fe(ll) concentration increased as the ascorbic
acid  broke through;  however,  Fe(ll) release continued near the
peak level of Fe(ll) release for  15 days beyond the ascorbic acid
breakthrough. In contrast, very little ascorbic acid broke through
and very  little Fe(ll) was released in the uncontaminated zone.
Some ascorbic acid may have been oxidized by oxygen, although
no significant changes  in the oxygen content were observed.
Released Fe(ll) may have been re-adsorbed by aquifer grains or
by released  colloids, in  which case  the Fe(ll) would promote
destabilization  and  deposition. Based  on these  results, it is
difficult to assess the dependence of PRD1 and silica colloid
recovery by ascorbic acid addition.
     Colloid Mobilization by Ionic Strength Decrease. Decreases
in groundwater ionic strength have frequently caused substantial
colloid mobilization  and  permeability reduction  during artificial
recharge [Nightingale and  Bianchi,  1977] and secondary  oil
recovery [Khilar and Fogler, 1984]. When  low ionic strength water
replaces high ionic strength water in aquifers with substantial clay
contents, an expansion of double  layers  leads to an increase in
electrostatic repulsion between colloids  and  grains and colloid
mobilization. In the  ferric oxyhydroxide-coated sands at Cape
Cod, however, we hypothesized that a decrease in ionic strength
would not cause substantial colloid mobilization because most of
the colloid-grain  interaction is attractive  (negatively charged
colloids, positively charged grain coatings). A decrease in ionic
                                                         16

-------
strength and expansion  of double layers would only serve to
strengthen this attractive  interaction. As  hypothesized,  the
decrease in ionic  strength  caused minimal  mobilization of the
natural colloids and the silica-coated zirconia colloids. There was
one exception  - when the  deionized water pulse followed the
0.1 M calcium chloride pulse, a substantial recovery of the silica-
coated zirconia was observed. In this extreme case, we surmise
that colloids were mobilized  where they were bound to  bare
quartz surfaces. No mobilization occurred without the calcium
chloride pulse preceding the deionized water pulse because the
contrast in ionic strength  was not sufficient.
Laboratory Experiments
Purpose
     To address one of the overall objectives of the  project, to
examine the dependence of colloid transport  and mobilization on
chemical perturbations, a set of laboratory experiments were
conducted to assess the rate of natural colloid mobilization from
the  ferric oxyhydroxide-coated quartz sands from  Cape Cod
under more controlled conditions. These laboratory experiments
included the following tests:
•  measurement of colloid mobilization  in the laboratory from
   undisturbed, oriented  sediment samples;
•  quantitation of the rate of colloid mobilization under controlled
   conditions;  and,
•  examination of a  range of chemical perturbations wider than
   possible in  the field to  understand the dependence of the
   amount and rate of  colloid  mobilization on the  chemical
   conditions.
To perform these experiments, we  developed a special column
packing technique to load the Cape Cod sediments into a column
with a minimum of disturbance to the grain-grain arrangement.
We then subjected these packed columns  to a  wide range of
chemical perturbations and measured colloid mobilization using
flow-through meters and  a data acquisition system.
Materials and Methods
     Sediment Characterization. Experiments were conducted
on uncontaminated and contaminated sediment samples from
the U.S. Geological Survey's Cape Cod field  site. To confirm the
location of the sewage plume in the sediment cores, pore waters
were  displaced with nitrogen pressure  from the  cores and
measured pH, specific conductance, and  dissolved oxygen.
Owing to some unavoidable atmospheric contact, specific
conductance was the best measure of contamination. Cores with
a specific conductance of less than 100 S cm"1 were identified as
uncontaminated and cores with a specific conductance of greater
than 300  S cnr1  were identified as  contaminated.  Cores of
intermediate specific conductance were not used in the laboratory
experiments.
     Column Packing and Setup. To  prepare  undisturbed,
oriented sediment columns, a stainless steel column (5.0 cm
length, 2.5 cm diameter) was gently rotated into a sediment core
through a hole drilled horizontally in the acrylic core sleeve. From
a hole drilled on the opposite side of the core,  the sediments were
lightly pressed into the column using a plunger of 2.5 cm diameter
[Ard, 1997]. The sediment was secured in the  column by threaded
stainless steel caps, Teflon® washers, a 0.2 |am stainless steel frit
on the influent end (bottom), and a 105 jam polypropylene mesh
on the effluent end (top). The dry weight of sediment packed into
the column averaged 38.6±2.0 g. The average column porosity
was 0.41±0.02 assuming a sediment grain density equal to that
of quartz  (2.65 g  cm"3).  The average column  pore volume,
9.0±0.5 mL, was measured bythe decrease in mass afterremoving
water from a saturated column and drying at 105°C overnight.
     Influent solutions were pumped through the column at a
flow rate of 0.15 mL min"1 (representing an interstitial groundwater
velocity of 0.7 md"1)  using syringe pumps to ensure constant
surge-free flow. The column effluent flowed through (1) micro-
cells measuring specific conductance and pH (volumes of 17 and
11,uL, respectively) and (2) a turbidity cell (volume about 2 mL) to
a fraction collector. A data acquisition system was used to record
data at 1 min intervals.
     Column Procedures. After packing each column, the
sediment was flushed  with a  high  ionic  strength solution
(0.5 M NaCI) to remove  colloids loosened by column packing
until a low background turbidity (roughly 5 NTU) was observed
(typically 5  pore volumes). Next, a low ionic strength solution
(5.0x10"4M  NaCI) representative of the  ionic  strength  of the
uncontaminated groundwater, was run through the column until
a low baseline turbidity (roughly 5  NTU) and specific conductance
was observed (typically 25 pore volumes).
     The sediments  were subjected to a series of chemical
perturbationsthatenvelopedthosetested inthefield experiments.
In some experiments, a single sediment column was subjected to
a sequence of chemical perturbations (Table 13).  In  these
experiments, each  influent solution was separated by a
"groundwaterflush"of5.0x10"4 M  NaCI run until the turbidity in the
effluent returned to the background level.  In other experiments,
a series  of individual columns  were subjected to  a series of
chemical perturbations (Table 14). The surfactant concentration
perturbations  ranged from one-tenth the value of the  critical
micelle  concentration (CMC)  for sodium  dodecylbenzene-
sulfonate, 1.2 mM[MukerjeeandMysels, 1971], to approximately
ten  times the CMC in a 5% solution.
     Colloid Characterization.  The elemental  composition,
morphology, mineralogy,  and zeta potential of the natural colloids
were characterized as described in the field experiments. The
total mass of mobilized colloids was determined by integrating
the  colloid concentration  in the column  effluent over the time of
release and multiplying this by the flow rate. This method assumes
deposition of mobilized colloids is negligible.
Results
     Colloid Characterization.  The elemental  and mineral
composition, morphology, and zeta potentials of the mobilized
colloids closely matched those of the colloids mobilized  in the
field experiments.  No significant differences were observed
between colloids mobilized by different water chemistries.
     Sequential Perturbation Experiments. The sequential pH
increase experiments revealed that very little colloid mobilization
occurred until the influent pH solution reached pH 9.5 and 10.5.
Only the two highest NaDBS concentrations  caused significant
colloid release. The ascorbic acid injections caused the most
release during the intervening groundwater flushes.
     At the highest  colloid mobilization rates by all  three
perturbations, it became apparentthatthe sequential perturbations
were removing too many colloids for the later perturbations to
accurately measure the rate and amount of colloid mobilization.
We were able to compare the total amount of colloids mobilized
to  assess  the  contrast between the uncontaminated and
contaminated  sediments (Table 15). During  pH increase
sequence, about twice as many colloids were  mobilized from the
contaminated sediment as in the uncontaminated sediment (in
contrast to the field experiments). During the NaDBS increase
sequence, about 7 times  as many colloids were mobilized in the
contaminated sediment as in the uncontaminated sediment (in
agreement with the field experiments). During the ascorbic acid
                                                         17

-------
Table 13.   Influent solutions for the sequential column colloid mobilization experiments on uncontaminated and contaminated
            sediments. Each of these influent solutions was followed by flushing by a "groundwater" solution of 5x10~4M NaCI
            (pH 6 and conductivity 70 uS crrr1).
             perturbation     constituents
unit
sequence of perturbations
elevated pH
(NaOH)
elevated
surfactant
(NaDBS)
elevated
reductant
(AscAc)
pH
NaBr
conductivity
NaDBS
NaBr
conductivity
AscAc
NaBr
conductivity
mM
mS crrr1
mM
mM
mS crrr1
M
mM
mS cnr1
5.5
1.5
0.20
0.12
1.5
0.20
0.0001
1.5
0.20
7.5
1.5
0.20
0.60
1.5
0.20
0.001
1.5
0.25
9.5
1.5
0.20
1.2
1.5
0.35
0.01
1.5
0.80
10.5
1.5
0.25
29
1.5
1.5
0.1
1.5
6.0
11.5
1.5
1.0
140
1.5
6.5



12.5
1.5
10






Table 14.   Influent solutions for the individual column colloid mobilization experiments on uncontaminated sediment.
perturbation
elevated pH
(NaOH)
constituents
pH
NaBr
conductivity
unit
mM
mS cm"1
sequence of perturbations
10.5
1.5
0.25
11.5
1.5
1.0
12.5
1.5
10
13.1
1.5
80
Table 15.   Total mass of natural colloids mobilized from the uncontaminated and contaminated Cape Cod sediments by
            sequential chemical perturbations. The mass of colloids mobilized in the pH and NaDBS experiments was calculated
            as the sum of colloids mobilized only during the chemical perturbation steps. For the AscAc experiment, the mass of
            colloids mobilized also included the colloids mobilized during the intervening groundwater flushing steps.


                                                   mass of colloids mobilized
                  perturbation        aquifer zone               (mg)
elevated pH
NaOH
elevated surfactant
NaDBS
elevated reductant
AscAc
uncontam
contam
uncontam
contam
uncontam
contam
3.4
7.2
0.14
1.0
0.47
0.89
                                                         18

-------
increase sequence, about twice as many colloids were mobilized
in the contaminated sediment as in the uncontaminated sediment
(in agreement with the field experiments).
     IndividualpH Increase Experiments. The amount of colloids
mobilized from the individual columns increased with increasing
pH up to pH 12.5, and then decreased substantially at pH 13.1
(Table 16; Figure 6). The pH 10.5 influent produced a long, slow
colloid release that lasted about 120 pore volumes before returning
to the background turbidity level. The pH 11.5 influent produced
greater and more  rapid  colloid release than pH  10.5; only about
25 pore volumes were required for return to background turbidity.
Colloid release at pH 12.5 influent was even greater and more
rapid than the lower pH  influents. The pH 13.1 influent produced
the most rapid colloid  release,  but the total mass  of colloids
released was small. Details of this experiment will  be described
in Ryan, Ard, Magelky,  and Elimelech [in preparation].
Discussion
     Colloid-Grain Morphology and Composition. The sediments
at the Cape Cod site are primarily composed of iron oxyhydroxide-
coated quartz [Coston  et al.,  1995]. At the ambient pH of the
groundwater (5.4 to 6.0), quartz is negatively charged (pH^ of
approximately 2.0 [Parks, 1967]). The colloids,  mainly clay
particles, also possess net negative charge at this pH value. Any
ferric oxyhydroxides present in the  coatings, however,  are
positively charged at the ambient pH (e.g., hematite, pH^ 6.7;
goethite, pHpzc8.5; lepidocrocite,  pH^ 7.4 [Parks, 1967]). On the
basis of these surface properties, it  is likely that the ferric
oxyhydroxide coatings  act as a cementing layer  between the
grains and colloids. Ryan and Gschwend [1990] made the same
argument for ferric  oxyhydroxide-coated quartz sands in two
Atlantic Coastal Plain sediments.
     Effect of pH on  Colloid Mobilization.  Substantial colloid
mobilization  began to occur at pH 9.5, but not at  pH 7.5. This
threshold appears to represent the pH at which the surface
charge of the ferric oxyhydroxide cement was reversed from
positive to negative. At pH 9.5, all of the mineral components of
the system  become negatively charged and  colloid release
occurs.  Ryan and Gschwend [1994] observed colloid release
threshold at a similar pH value during colloid release experiments
in a ferric oxyhydroxide-coated quartz sand from the Cohansey
Formation in New Jersey.
     As the influent pH was increased to 10.5,11.5,  and 12.5, the
rate and total mass of colloid release increased. The increase in
pH increased the negative zeta potentials on the  colloid and grain
surfaces. The increased negative zeta potentials caused greater
colloid-grain repulsion. Relating the rate and amount of colloid
release to the amount of  repulsion  is challenging for natural
systems, but we have observed a clear relationship here between
the rate and extent of colloid  mobilization and the colloid and
grain surface charge.
     Increasing the influent pH to 13.1, however, substantially
decreased the amount of colloid release, while the rate of release
remained rapid.  Similar colloid  release  behavior has been
observed by Kolakowski and Matijevic [1979] in a model system
of chromium  hydroxide colloids and glass beads. During the
transition from ambient pH to 13.1, the colloid and grain surfaces
became increasingly repulsive; however, as the column effluent
pH reached 13.1, the ionic strength exceeded 0.1 M. The high
ionic strength compresses the repulsive double layers and shields
the colloid and grain from repulsion; hence, colloid release was
shut down. The  initial pulse of colloid release was  still rapid
because the pH in the column was  rapidly increasing  to
13.1 through the pH range that promoted colloid release. At some
pH value between 12.5 and 13.1, the colloid and grain surfaces
have  reached their maximum negative  surface  charge. The
additional increase in pH is only raising ionic strength, not making
the colloid-grain interactions more repulsive. Inhibition of colloid
release by elevated ionic strength was also observed in the field.
Although a large release of colloids was measured at the 1 m
transport distance during the pH 12.5 injection, little release was
observed at the injection point itself.
     In contrast with the sequential  pH increase experiments,
the release of colloids in the uncontaminated zone was much
greater than that in the contaminated zone during the  individual
column experiments. The individual column experiment results
agree  with the  field experiments. Greater  release in the
uncontaminated zone can  be attributed to its lower buffering
capacity.
     Sediment  Disturbance  and  Orientation.  Employing
undisturbed and oriented samples forcolumn experiments makes
this laboratory work more applicable to in situ systems.  Using
disturbed Cape  Cod  sediments,  Roy  and Dzombak [1996]
measured greater  amounts  of colloid  release  under similar
chemical conditions. Disturbing the sediments through column
packing can  loosen  attached colloids and destroy  sediment
arrangements that may have an effect on colloid mobilization. If
repacking  is  avoided and  columns are  sampled without
disturbance  but  the original orientation of the  sediments is
Table 16.   Results of the elevated pH colloid mobilization experiments for uncontaminated sediment in the individual column
            experiments.
pH
10.5
11.5
12.5
13.1
peak colloid
concentration
(mg I/')
14.5
138
196
314
time of
peak release
(pore volume)
26
3.3
2.6
1.4
time to return
to background
(pore volume)
94
20
11
4.3
mass of colloids released
(mg)
4.4
6.1
8.4
0.78
                                                         19

-------
   13 -

   12 -

   11 -

   10 -

   9 -

   8 -

   7 -

   6 -
                                            (a)
PH
conductivity
                    40     60      80
                       pore volume
co
,§
|-
I
I
                                               120
                                                                     40 -
                                                                      0 -
                                                           13-1

                                                           12 -

                                                           11 -
                                                                                                    (b)
                                                                           0       5       10      15
                                                                                          pore volume
                                                              r5
                                                              - 4
                                                                                                          hi  1
                                                                                                           - 0
                                            (c)
                                                                                                    (d)
200 -

'-> 150 -
en
E,
-S 100 -
'o
8
50 -
0 -
13 -

12 -
11 -
10 -
I
Q.
9 -


8 -
7 -


«
1

•
*
•1
* \
" ^^kA

• pH
* conductivity



»


JC
j
- 10

-8 'P
o
-6 I
>,
-1-'
-4 -S
o
3
•D
-* §
n
0 10 20 30 40 50 60
pore volume
                                                                        350

                                                                        300

                                                                        250

                                                                        200
                                                                         13

                                                                         12

                                                                         11

                                                                         10
                                                                                            4       6
                                                                                           pore volume
                                                                                                              ro

                                                                                                              t
                                                                                                                   -20
Figure 6.    Natural colloid mobilization from oriented, undisturbed columns of uncontaminated sediment by pH elevation: (a) pH 10.5
             influent, (b) pH 11.5 influent, (c) pH 12.5 influent, and (d) pH 13.1 influent.
                                                              20

-------
changed, colloid mobilization results could still be very different
from those witnessed in an in situ experiment. Sediment grains
arrange themselves relative to the flow of groundwater around
them, if the sediment orientation is ignored then an influent would
be approaching the sediments at a different angle and towards a
different sediment face than had previously been exposed to
approaching groundwater. Though repacked and disoriented
columns can help define colloid mobilization and transporttrends,
undisturbed and oriented columns will provide more applicable
information.
Modeling
Purpose
     To address the third of the major objectives of this project,
the development of a colloid transport model that would describe
the colloid transport in a contaminant plume, we undertook the
modeling of colloid transport in a geochemically and  physically
heterogeneous porous medium similar to that found at the U.S.
Geological Survey Cape Cod field site.  As  a first step toward
modeling the effects of a contaminant plume on colloid transport,
this model simulates the effect of heterogeneity and the dynamics
of colloid deposition. The original objectives of developing a
colloid transport model capable of simulating all of the processes
playing a role in the transport of colloid in a contaminant plume
were  not met during this  research period. Ongoing  model
development is still striving towards those original objectives.
     Field and laboratory investigations on the transport of
colloidal particles in aqueous porous media have demonstrated
that  advection,  hydrodynamic dispersion, particle deposition,
and  particle  release  are the primary mechanisms controlling
colloid transport in porous media [Harvey et al., 1989; Tobiason,
1989; Elimelech and O'Melia, 1990; Elimelech, 1991; Lindqvist
and  Enfield, 1992;  Higgo  et al., 1993;  Harvey et al.,  1995;
Kretzschmar et al., 1995;  Penrod et al.,  1996; Harmand  et al.,
1996; Pieper et al., 1997]. Various models involving these four
main colloid transport mechanisms have been developed based
on colloid mass balance overa representative elementary volume
(REV) of a  porous  medium. Among these  four main  colloid
transport mechanisms, colloid deposition has been investigated
more extensively.
     The kinetics  of particle deposition have  been  derived
theoretically and  measured experimentally.  The theory of
Derjaguin-Landau-Verwey-Overbeek(DLVO) was generally used
to describe the surface-surface interactions to predict the particle
deposition or  release rates [Elimelech  and O'Melia,  1990].
However, theoretical predictions are generally several orders of
magnitude  greater than experimental  observations  under
unfavorable  deposition conditions (i.e.,  when repulsive double
layer  interactions  predominate). Various attempts have been
made to explain  this discrepancy, including  attachment in
secondary energy minima [McDowell-Boyer,  1992; Stumm and
Morgan, 1995], interfacial dynamics of double layer interaction
[Elimelech and O'Melia, 1990], and surface non-idealities [Song
et al.,  1994].  Among  these  explanations,  geochemical
heterogeneity (surface charge heterogeneity) is considered to be
the most probable cause of the anomalous colloid deposition
rates  observed in porous media [Song et al.,  1994].
     The geochemical heterogeneity of granular porous media
was modeled either as random microscopic sites or as patches
[Song et al.,  1994]. Patchwise  charge heterogeneities  are
ubiquitous in subsurface environments  due to geochemical
variabilities  inherent  in  mineral grains.  Johnson et  al.  [1996]
adopted the  patch model to describe  the  geochemical
heterogeneity  of  porous  media. They  assumed a constant
geochemical heterogeneity for the entire porous medium and
incorporated it into a colloid transport model.
     The dynamics of colloid deposition have also been studied.
A linear Langmuirian blocking function was proposed by Privman
et al. [1991]. Song and Elimelech [1994] extended the model by
including the non-uniform deposition  resulting from  spherical
collector geometry and surface heterogeneities. The  nonlinear
random sequential adsorption (RSA) model was employed by
Johnson and  Elimelech [1995]  to describe the dynamics of
blocking in colloid deposition. More recently,  Johnson  et al.
[1996]  have incorporated the RSA model for dynamic blocking
process and the patchwise model for geochemical heterogeneity
into a  colloid  transport model.  They found good  agreement
between model predictions and the experimental measurements.
     There are relatively fewertheoretical formulations for colloid
release, advection, and hydrodynamic dispersion.  The  colloid
release is usually described as a kinetic process (first-order
mechanism) instead of a dynamic process in colloid transport
equation [Chrysikopoulos et al., 1990]. When the hydrodynamic
chromatography effect is  not obvious in  colloid transport,  the
difference between the colloid advection velocity and the solute
advection velocity (or colloid and solute dispersion coefficient) is
usually ignored, and the hydrodynamic dispersion linearly depends
on the  advection velocity for both colloid and solute transport.
     Most of the developed models only describe colloid transport
in  physically  homogeneous porous  media. However, porous
media in subsurface environments are physically heterogeneous
[LeBlanc,  1984; Hess et al., 1992]. Only Abdel-Salam and
Chrysikopoulos [1995] modeled colloid transport in a fractured-
rocks matrix using lognormally distributed fracture aperture. The
physical heterogeneity was described as a random field.  Saiers
et  al. [1994] carried out experiments of colloid transport in a
structured-heterogeneous porous medium. In these experiments,
the column was packed with different sand layers, which were
parallel to  the flow  direction;  each  layer was packed
homogeneously. This physical heterogeneity can be viewed as
layered distributed. The results  of both  studies showed that
physical heterogeneity of porous media can affect colloid transport
significantly. These researchers focused on colloid advection
and hydrodynamic dispersion to explain the consequences of the
heterogeneous flow field on colloid transport. Although the flow
velocity may also affect the  kinetics of particle  deposition or
release, such effects were not considered in the above studies.
     The previous modeling of colloid transport in geochemically
heterogeneous porous media only dealt with the porous medium
with a constant geochemical heterogeneity. However, a spatially
distributed geochemical heterogeneity is very likely to exist in a
porous medium due to the inherent variability of minerals. Thus,
it is  important to study the role  of a spatial  distribution of
geochemical heterogeneity in colloid transport. Since geochemical
heterogeneity has only recently been introduced to colloid transport
studies, there have been no attempts to incorporate a distribution
of geochemical heterogeneity into colloid transport modeling.
     We developed a two-dimensional model forcolloid transport
in physically and geochemically  heterogeneous porous media
[Sun, Elimelech, Sun, and Ryan, submitted]. In this  model, a
patchwise geochemical heterogeneity and dynamic aspects of
particle deposition and release are included in the governing
equations  for colloid transport, which are coupled with the flow
equation.

Model Development
     We consider a confined aquifer, where the fluid follows a
steady  laminar motion and  the  suspended colloidal particles
                                                         21

-------
travel at the fluid velocity. The colloidal particles are assumed to
be Brownian (i.e., less than about 1|am) and monodisperse. The
steady state flow field is derived from the transient flow equation
and  incorporated  into the colloid transport equation.  Spatial
distributions of the physical and geochemical heterogeneities of
the subsurface porous medium are rigorously incorporated in the
model.
     Flow Field. The transient  flow equation for a  fluid in  a
confined subsurface porous medium,  such as a  confined
groundwater aquifer,  is  usually written as
where  h is the hydraulic head, t is the time, St is the specific
storage, K is the hydraulic conductivity, and O is the pumping or
recharge rate. Under natural gradient flow  conditions, colloid
advection  can be described by the steady-state flow equation.
The spatially distributed hydraulic heads are used to calculate the
velocity field by applying Darcy's law
        q=-K-W?
(3)
where  h is the hydraulic head gradient and q is Darcy's velocity.
The  average pore velocity (V), which  is used  in the colloid
transport equation, is the ratio of Darcy's velocity to porosity.
     Physical Heterogeneity of Subsurface Porous Media. The
spatial variation of hydraulic conductivity is the  principal cause of
heterogeneous flow field that further influences colloid transport
and the resulting particle concentration in the porous medium.
Two types of physical heterogeneity are investigated, namely,
layered heterogeneity and  random heterogeneity.
     In a layered, physically heterogeneous subsurface porous
medium, the porous medium is made up of several homogeneous
layers. Thus, while each layer is homogeneous (i.e., with constant
hydraulic conductivity), the entire system is heterogeneous.
Porous media with large blocks of macropores or fractures may
be described as layered heterogeneous.
     Substantial progress has been made in the pasttwo decades
to understand the random physical heterogeneity of groundwater
aquifers. Evidence  from  field-scale hydraulic  conductivity
measurements indicates that the spatial distribution of hydraulic
conductivity is lognormal [e.g., Freeze, 1975; Hoeksema and
Kitanidis, 1984; Sudicky, 1986; Hess, 1989]. It was also found
that there exists a non-Gaussian behavior of the log-transformed
hydraulic conductivity at relatively small scales, and thatthis non-
Gaussian behavior shifts to Gaussian behavior  as the length
scale increases [e.g., Painter, 1996; 1997; Liu and Molz, 1997].
Because of lack of knowledge on this transition length scale and
the fact that lognormally distributed hydraulic conductivity has
generally been used  by numerous hydrologists [e.g., Gelhar et
al.,  1979; Gelhar and Axness, 1983; Dagan, 1984; Bellin etal.,
1992],  a lognormal distribution is adopted here to describe the
random spatial variation of hydraulic conductivity.

     Let Y=lnK, with a constant mean myand variance (TY .The
covariance function  of  Y is  assumed to have  an  isotropic
exponential form,
        Cy(r)=or2exp -
(4)
where rY is the planar distance vector between two positions in
the heterogeneous domain and /yisthe integral scale of V. Using
statistical properties of the spatial distribution, the random field of
hydraulic conductivity can be generated by the turning band
method [Mantoglou and Wilson,  1982; Tompson et al., 1989].
     Geochemical Heterogeneity of Subsurface Porous Media.
Coatings and patches of oxyhydroxides (iron and aluminum) on
subsurface mineral grains are the main source of  geochemical
heterogeneity in groundwater aquifers [Coston etal., 1995; Ryan
et al., 1999]. These coatings on  mineral grain surfaces provide
favorable sites (area) for colloid deposition.  Here, we adopt the
patch model [Song et al.,  1994] to describe the  geochemical
heterogeneity of subsurface porous media.  The model is
characterized by the heterogeneity parameter, A,, which is defined
as the ratio of the surface area favorable for colloid  deposition to
the total interstitial surface area over a REV of a porous medium.
The surface area favorable for deposition is usually characterized
by a surface charge opposite that of the colloids. Because most
colloids are negatively charged, the favorable deposition areas
are patches of positively charged minerals (e.g., iron and aluminum
oxides, clay edges).  Note that colloid deposition or release can
occur on both the favorable and  unfavorable fractions, albeit at
much different rates.
     Because the chemical composition of subsurface minerals
and  solution  chemistry  vary  spatially in  subsurface aquatic
environments,  the geochemical  heterogeneity, defined over a
REV, may vary significantly throughout the subsurface porous
medium. The geochemical heterogeneity of a  porous medium
can be assumed to be constant over the entire porous medium,
or to have  different values at different locations in the porous
medium. Accordingly, two  spatial  variations of  geochemical
heterogeneity are considered: layered geochemical heterogeneity
and random geochemical heterogeneity.
     Compared to layered geochemical heterogeneity, detailed
statistical information on the chemical properties ofthe subsurface
porous  medium  is  needed to  model  random geochemical
heterogeneity. To date,  there are no reported  studies on the
random field of geochemical heterogeneity of subsurface porous
media in relevance to colloid transport. Several studies on solute
transport in heterogeneous porous media have described the
variation of solute sorption coefficients by a normal distribution
[e.g.,  Black and Freyberg, 1987; Chrysikopoulos  et al., 1990;
Bosma and van derZee, 1993]. Although solute transport behavior
is quite different than colloidal transport behavior, we adopt a
similar approach and describe the random field of  geochemical
heterogeneity as normally distributed with a constant mean E(X.)

and a variance <7\. The turning band method is used to construct
the  two-dimensional random field  of normally distributed
geochemical heterogeneity,  with a first-order  exponential
autocorrelation function:
                                                             (5)
      where L  is the integral scale of
     The  average value of the geochemical heterogeneity
parameter in groundwater aquifers is usually thought to be small,
on the order of a few percent [Heron etal., 1994; Kretzschmaret
al., 1995; Coston et al., 1995; Ryan et al., 1999]. With a mean
value of only a few percent, normal distribution cannot cover a
wide range of geochemical  heterogeneity. Thus, in addition to
normal distribution, a lognormal distribution will be used when
significant variations of geochemical heterogeneity, with a small
mean value of only a few percent, are desired.
                                                         22

-------
     Colloid Transport Equation. The colloid transport equation
can be derived from mass balance of colloids over a REV of a
porous medium. There  are three main mechanisms controlling
colloid transport: hydrodynamic dispersion, advection, and the
colloid exchange between the stationary  solid  matrix and the
mobile colloidal phase through colloid deposition and release.
These mechanisms can  be described  by the generalized
advection-dispersion equation [e.g., CorapciogluandKim, 1995]:
                                                       (6)
where  C is the mass concentration of colloids in the aqueous
phase, S is the ratio of the colloid mass captured by the solid
matrix  to  the  total mass of solid matrix,  D is the particle
hydrodynamic  dispersion coefficient, Vis the particle velocity, e
is the porosity of the porous medium, and pb is the  bulk density of
the porous medium. Because the average pore radius in sandy
aquifers  is quite large compared to the size  of Brownian
(submicrometer-size) colloidal particles,  size  exclusion effects
are not considered. Thus, the particle velocity and  interstitial fluid
velocity are assumed to be equal. Similarto the relations developed
for solute dispersion, the particle dispersion coefficient is linearly
dependent on the interstitial velocity:
                —                VV
        Dij=aLV8ij+(aL-aT)^I+DdT8ij
(7)
where  V. is the component of the spatially distributed interstitial
velocity along  direction /, Dd is the colloid diffusion coefficient
obtained from the Stokes-Einstein equation, aL and  oc^are the
longitudinal and transverse dispersivities, respectively, and 7 is
the porous medium tortuosity.
     To appropriately describe the dynamic aspects of colloid
deposition or release, the mass transport equation should be
expressed in terms of colloid number concentration rather than
mass concentration [Johnson and  Elimelech, 1995; Johnson et
al., 1996], that is

                                                       (8)
where n is the number concentration of colloids,  8 is the fractional
surface coverage,  defined as the total cross section area of
deposited colloids  per interstitial  surface area of the porous
medium solid matrix,/ is the specific surface area (i.e., interstitial
surface area per porous medium  pore volume), and a  is the
radius of colloidal particles. It can be readily shown that Eqn. 8 is
equivalent to Eqn. 6.
     Colloid Deposition and Release.   Using the patchwise
model  for  geochemical  heterogeneity,  the particle surface
coverage rate of mineral grains is given by [Johnson etal., 1996]
                                                       (9)
When considering the dynamic aspects of particle deposition and
release, the rate equations corresponding to the favorable and
unfavorable surface fractions can be expressed as
                 Qt
                                                           (10a)
                                                           (10b)
       where the subscripts/and  «  represent the favorable (A) and
       unfavorable (1-X) REV surface fractions, respectively, kdep is the
       colloid deposition  rate coefficient,  k^ is the first order colloid
       release rate coefficient,  and B(0) and  R(9) are the dynamic
       blocking and releasefunctions, respectively. The colloid deposition
       rate coefficient  is related to the single collector efficiency t|
       commonly used in filtration theories as [Elimelech et al., 1995]
                    _riV_ari0V
                dep"4e"  4e
                                                     (11)
where  V is colloid advection velocity, e is the porosity of the
porous medium,  a is the collision  efficiency, and t|0 is the
favorable single collector removal efficiency.
     The dynamic blocking function B(9) describes the probability
of a colloid contacting a portion of collector surface unoccupied
by previously deposited colloids [Song et al., 1994]. It accounts
for the blocking effect of deposited colloids on the  particle
deposition rate. Two types of dynamic blocking functions are
generally recognized: Langmuirian dynamic  blocking function
and  random  sequential adsorption  (RSA) dynamic  blocking
function.  Recent experimental  investigations  have shown that
the RSA model describes the dynamics of particle deposition in
porous media much  better than the  conventional Langmuirian
model [Johnson and  Elimelech, 1995; Johnson et al., 1996].
     The general form of the RSA dynamic blocking function is
[Adamczyket al.,  1992]
                                                            (12)
       where emax is the maximum attainable surface coverage, a, and
       a2, a3  are coefficients that can be found theoretically (for ideal
       particles and collector surfaces) or empirically. The coefficients
       used by Johnson and Elimelech [1995] for B(9) will be used in this
       colloid transport model as they were found adequate to describe
       the dynamics  of blocking  in flow of monodisperse  latex
       microspheres in columns packed with spherical  and uniform
       glass beads [Johnson and Elimelech,  1995]. Because colloid
       deposition onto the favorable surface fraction is usually irreversible,
       the RSA model can be used to describe the dynamics of particle
       deposition onto the favorable surface fraction. A similar dynamic
       blocking function was also chosen to describe the blocking of the
       unfavorable fraction, although the deposition onto the unfavorable
       surface fraction  was assumed to be reversible with a non-zero
       release rate. This assumption, however, has negligible effect on
       the colloid transport behavior since the deposition rate on the
       unfavorable surface fraction is much smallerthan on the favorable
       fraction,  and the maximum surface coverage for the unfavorable
       surface  fraction is much  smaller than the maximum surface
       coverage on the favorable surface fraction.
           Somewhat analogous to the dynamic blocking function, the
       dynamic  release function describes  the  probability  of colloid
                                                          23

-------
release from porous media surfaces covered by retained colloids.
In principle, this function should depend on the colloid residence
time and the retained colloid concentration [Johnson et al., 1996].
When R(0) = 1, the release terms in Eqn. 10 represent first order
kinetics release mechanism. Because at the present  time the
mechanisms of colloid release are poorly understood, only a first
order release rate will be used in this paper.
     Correlation between  Physical Heterogeneity and Colloid
Deposition Rate.  In modeling colloid transport, the variation  of
flow field will change the colloid concentration distribution in the
studied domain not only by affecting hydrodynamic dispersion
and advection, but also by influencing the colloid deposition rate.
For Brownian colloids where deposition rate  is controlled by a
convective-diffusive mechanism [Elimelech et al.,  1995] there
exists a positive relationship between the hydraulic conductivity
and the particle deposition rate. Based on  Eqn. 11, the colloid
deposition rate (kdep) is proportional to (r|0V), with t|0 for Brownian
colloids being proportional to V2'3 [Elimelech etal., 1995]. Hence,
combining this relationship with Darcy's law one obtains that the
colloid  deposition rate (kdep) is proportional to K"3. A consequence
of this relationship is that a random field of hydraulic conductivity
leads to a random field  of colloid deposition rate as well.
     We assume that P(x) = In k^x) is normally distributed with

a mean 

and variance cr2p, and has a similar form of the covariance function as the hydraulic conductivity field. To describe the correlation between the random hydraulic conductivity field and the colloid deposition rate, it is further assumed that (13) where coandyare correlation coefficients and Y' is the perturbation of the hydraulic conductivity field. When co =0 and y>0, Pand Y are perfectly positively correlated; when o>=0 and y<0, Pand Y are perfectly negatively correlated; and when co^O, Pand Yare not perfectly correlated. For Brownian colloids, w=0andy>0; the value of Y is chosen as 1/3, for the reason discussed above. Numerical Procedures. In the colloid transport model presented in the previous section, the transient flow equation is coupled to the colloid transport equation. Numerical solution can be obtained with both transient and steady state flow fields using the multiple cell balance (MCB) method [Sun, 1995]. The flow region in our model is a vertical rectangular domain, with the horizontal x axis ranging from 0 to 3 m and the vertical z axis ranging from 0 and 1 m. The computational domain Q is encircled by the line boundary F. Initial and Boundary Conditions. For the flow equation, the initial and boundary conditions for the flow domain are specified as follows: h(3,Z,t) = for f>0, (3,Z)GT4 (14e) h(x)=h0 at for t>0, (0,z)er, (i4a) (14b) dh(x,z,t) = 0 for f>0, (x,0)er2 (14c) Z=0 = 0 for t > 0, (x,l) e T3 where F = Fj UT2 LIT, UT4 , and h1 and h2 are fixed values of hydraulic heads on the boundaries. The initial and boundary conditions forthe colloid transport equation are specified as follows. Initially the porous medium has no deposited colloids (i.e., zero surface coverage, 9= Ou=0). At the four boundaries (T ), zero dispersive flux boundary conditions are specified. Furthermore, a given concentration of colloids is injected into the domain at t>0. The type of colloid injection can be classified as pulse injection or continuous injection depending on the duration of the injection. The mode of injection can be characterized as point injection or line injection based on the number and locations of injection wells. The injection is set as the boundary condition forthe colloid concentration. Multiple Cell Balance Algorithm. Because there are no analytical solutions forthe flow and colloid transport governing equations, we adopted the multiple cell balance (MCB) method [Sun and Yeh, 1983] to solve the two-dimensional transport model numerically. The MCB method was originally derived for solving the coupled groundwater flow and solute transport equations for solutes experiencing equilibrium sorption [Sun, 1995]. The method has never been applied to the more complex problem of colloid transport. The details on the numerical formulation and procedures are given in the doctoral dissertation of Sun [Sun, 1998]. Validation of the Numerical Code. Previous studies suggested the MCB method provides adequate solution for two- dimensional non-reacting solute transport problems [Sun, 1 995]. We compared the numerical solution based on our MCB code for the transport of a tracer in a two-dimensional semi-infinite isotropic porous medium with the analytical solution provided by Leij and Dane [1990]. The numerical results closely agreed with the analytical solution. To validate the MCB code for colloid transport, the analytical solution derived by Lapidus and Amundson [1952] forthe one-dimensional solute transport problem with finite rates of sorption (/cr) and desorption (k2) was compared with our numerical solution forthe following problem: =L dt Ldx2 dt e dt (15) (16) The initial concentration is set at zero, the concentration at the inlet boundary is given as a constant C0, and the dispersive flux of colloid is set at zero to the outlet boundary. The analytical solution of Lapidus and Amundson [1952] is given by Vx where (14d) z=l (17) (18) ——id ;4Dr err 24


-------
and
                     4D
                                                      (19)
     As shown in Figure 7, the numerical results obtained from
the MCB code are in very close agreement with the analytical
solution.
Results and Discussion
     The newly developed 2-D colloidal transport model is used
to conduct a numerical investigation of colloidal transport in
physically and geochemically heterogeneous porous media. We
first illustrate the effect of key model parameters on the general
colloid transport  behavior.  This  follows by a systematic
investigation of colloid transport in layered as well as randomly
heterogeneous subsurface porous media.
     Influence of Key Model Parameters. The basic values of the
model  parameters and the range of their variation during the
numerical investigation are  listed in Table 17. The  range of
parameter values covers possible scenarios of colloid transport
in sandy aquifers, such as the glacial outwash  sandy aquifer in
Cape Cod,  Massachusetts, which has been used extensively in
field investigations [LeBlancetal., 1991;  Garabedianetal., 1991;
Hess et al., 1992; Gelhar et al.,  1992; Sun, 1995;  Harvey et al.,
1989; Ryan and Gschwend, 1990;  McCarthy  and Degueldre,
1993; Johnson etal., 1996; Pieperetal., 1997]. Colloidal particles
are introduced at the boundary x = 0 as a pulse injection with a
duration of 0.5 days. The results (Figure 8) are  presented as
relative colloid concentration n/n0 along the flow direction x at a
certain observation time (f = 0.75 d).
     Under the examined conditions, an increase in hydraulic
conductivity results in enhanced colloid migration and a wider
spreading of the colloid concentration profiles (Figure 8a). With
a constant hydraulic head gradient, a greater hydraulic conductivity
results in larger flow velocity so that the colloid advection velocity
increases.  Because  the particle  dispersion coefficient is
proportional  to the colloid  advection  velocity,  dispersion
(spreading) of colloids also increases asthe hydraulic conductivity
is increased. In addition, the capture of colloids traveling through
the porous medium decreases with increasing flow velocity, thus
resulting in a  slightly attenuated particle concentration profile.
     Hydrodynamic dispersion  is  also controlled  by  the
longitudinal and transverse dispersivities (Eqn. 7). The ratio of
longitudinal dispersivity to transverse dispersivity is typically in
the range of 5 to 20 [Sun, 1995]. We assumed a ratio of 5  and
investigated the effect of varying the longitudinal dispersivity.
Because longitudinal dispersivity is scale dependent, and our
problem is of local scale (ca. 3 m), only a narrow range of values
was selected for the longitudinal dispersivity.  The results
(Figure  8b) show that small changes in longitudinal dispersivity
lead to  relatively large changes  in the  colloid  concentration
profiles.
     To investigate the  effect of particle  deposition rate,  a
constant geochemical heterogeneity (X=0.01) was assumed. We
fixed the favorable particle deposition rate coefficient, and adjusted
the unfavorable deposition rate by choosing different values for
the collision efficiency of the unfavorable surface fraction au.  The
results  (Figure 8c) show that particle  deposition rate  can
substantially affectthe colloid concentration profile. Asthe collision
efficiency au increases, the colloid  deposition  rate on  the
unfavorable fraction increases, and less colloids can be detected
in the aqueous  phase. The magnitude of the collision efficiency
au  reflects the effect  of changes in the  solution chemical
composition.
     By fixing the particle deposition rate coefficients kdg.f and
kdepu, the overall particle deposition rate can be controlled by the
geochemical  heterogeneity parameter /L The marked effect of
geochemical heterogeneity on colloidal transport is illustrated in
Figure 8d. An increase in geochemical heterogeneity results in
increased overall colloid deposition rate and reduced concentration
of colloids in bulk solution. For the conditions investigated in
Figure 8d, a substantial geochemical heterogeneity of subsurface
porous media (>10%) may result in nearly complete immobilization
                               1.0
                              0.8
                              0.6
                              0.4
                              0.2
                              0.0
                                     0.0
                                             0.5
                                                     1.0
                                                              1.5

                                                            x(m)
                                                                      2.0
                                                                              2.5
                                                                                      3.0
Figure 7.    Comparison of numerical solutions (symbols) with analytical solutions (lines) for reactive solute transport in isotropic
             semi-infinite porous medium. The parameters used for this simulation are VDarc = 1.0 m s~1;  aL = 0.05 m,  aL: aT = 5:1;
             PeL = 1;  Courant No. = 0.25-1.0. The observation times from t1 to t6 are 0.5, 1.0, 2.0, 3.0, 4.0, 5.0 d.
                                                          25

-------
Table 17.    Basic values and ranges for parameters in the colloid transport model.

                 parameter                        value                   range

              hydrologic  parameters

                         Vh
              transport parameters
lO-2

102
   1CH-10-1

10°~103m/dav
Ss
ocL(m)
aL/aT
e
lo-4
0.05
5
0.4

0.01-0. 7m
5-20
0.3-0.5
dc (mm)
f a (m2/m3)
dp (u,m)
pp (g cm3)
Co(mgL-i)
n0 (# m3)
a
b
M%)
Wuc(md-')
kdep,fc(md-1)
kdetu (hr1)
w
0.3
30000
0.3 0.01-1
2.5
10
2.8xl014 IQH-IO15
] Q-3 10"4— 10°
0.0259
0.1 0.1-10
6.5xlO-6
6.5x10-3
o.o io-4~io-2
0.0 0.0
                         3 Determined from 3(1-e)/(e aj
                         b Determined from the method of Elimelech and Song [1992].
                         0 Determined from Eqn. 11.
of colloidal particles as shown by the flat,  attenuated  colloid
concentration profile.
     The colloidal transport model assumes that particle
deposition  onto  the favorable surface fraction is  irreversible;
hence, the colloid release rate from the favorable surface fraction
iszero. This assumption has been confirmed in particle deposition
studies involving oppositely charged  particles  and  collector
surfaces [Elimelech et al.,  1995]. On the other hand, in particle
deposition  studies  involving similarly charged  particles and
collector surfaces, a finite rate of colloid release can be detected
[Ryan and Elimelech, 1996]. Hence, we investigated the effect of
colloid release rate from the unfavorable surface fraction on the
colloid concentration profile as shown in Figure 8e. The  results
demonstratethat larger release rate coefficients result in increased
colloid concentration in  the aqueous phase, whereas smaller
release rate coefficients have no effect on the colloid concentration
profile. Since the colloid release rate depends on the concentration
of deposited particles (first order  kinetics), the effect of colloid
           release on  the  colloid  concentration profile depends on  the
           overall  colloid  deposition  rate onto the  unfavorable  surface
           fraction. Figure 8f demonstrates that the model solution is very
           sensitive to particle size. Particle size influences colloidaltransport
           mainly through its effect on colloid deposition rate. As expected
           for deposition of Brownian particles, which is  controlled by  a
           convective-diffusion  mechanism,  the deposition rate becomes
           smaller as particle size increases.  Consequently, larger particles
           migrate faster in the porous medium and their concentration in
           the liquid phase is greater than that of smaller particles.
                Colloid Transport in Layered Heterogeneous Porous Media.
           The porous medium was divided into three horizontal layers,
           parallel to the flow direction. The layers are denoted as layer I (0-
           0.3 m), layer II (0.3-0.7 m), and layer III (0.7-1.0 m) from bottom
           to top. Layers I and  III were assigned the same heterogeneity
           parameter  values, whereas a different parameter value was
           assigned to the  middle  layer II.  The colloid suspension was
           assumed to be fed continuously (line injection) into the porous
                                                           26

-------
   1   '.'I        X
  /   J\      /\
 f   /  k,   /    ''

 /   /    \/
 /  /      A
'  /     / \
             *  P   s'&

            && -a- -A
                                                             (a)
                                                                                                        ------ a
                                                                                                        "••  '"-  "
                                                                                                         -  (b)
I
                                      - -v - v - v -7-
                                                 I
                                                             (c)
                                                                            - -V- -V- -V -V" -7- -V -V- -V- -
                                                                                                            (d)
                                                             (e)

                                                                                                            (f)
 Figure 8.    The role of model parameters in colloid transport in physically homogeneous porous media. The results are presented
              as relative colloid concentration n/ng along the flow direction xat a certain observation time (f = 0.75 d). The effects of
              (a) hydraulic conductivity, (b) longitudinal  dispersivity,  (c)  collision efficiency (deposition rates), (d) geochemical
              heterogeneity, (e) release rate from unfavorable surface fraction;  and (f) particle size on colloid concentration profile
              along z = 0.5 m. Parameter values are shown in Table 17.
                                                             27

-------
medium at the inlet boundary (x=0), with 11 injection points set at
0.1 mintervalsalongthezdirection. Observations of concentration
profiles over the entire two-dimensional porous medium domain
are  presented  for f=0.75 d.  The  physical and  geochemical
heterogeneity parameter values used in the  numerical
investigation (represented by  K and  1,  respectively) were
comparable to those reported for the Cape Cod sandy aquifer
[Leblanc et al.,  1991].
     The effect of layer-distributed  physical  heterogeneity on
colloid transport is illustrated in Figure 9. The hydraulic conductivity
of the middle layer(layer II) of the porous medium is twice as large
as the hydraulic conductivity in the layers above and below.
Therefore, the fluid flows in the central layer faster than the other
two layers, and  most of the colloids migrate with the flow through
the more permeable layer. This example points outthe paramount
importance of preferential flow paths in colloid transport.
     Because transverse dispersion reduces the  amount  of
colloids passing through  the preferential flow path, the role  of
longitudinal and transverse dispersivities was also investigated.
Two different ratios of longitudinal to the transverse dispersivities
(1 and 10) were  studied, as shown in  Figure 10. When the
transverse dispersion is relatively large (ccL/ccT=1.0), the extent of
preferential flow in the middle  (most permeable)  layer of the
porous medium is reduced (Figure 11 a). However,  when the
transverse dispersion is relatively small (otL/ocT=10.0),  the
preferential transport of colloids in the middle layer of the porous
medium is enhanced (Figure 11 b). The results demonstrate that
hydrodynamic dispersion can influence colloid transport in layered
heterogeneous porous media, but the effect is not strong enough
to explain the preferential transport of colloidal particles.
     The effect of a layered  geochemical  heterogeneity on
colloid transport in physically homogeneous (constant hydraulic
conductivity)  subsurface porous medium is shown in Figure 12.
The central  layer had a very  small value of  geochemical
heterogeneity (1=0.001),  whereas the upper and lower layers
had  much higher values (1=0.025). The results clearly show that
the increased particle deposition rate of colloids onto thefavorable
                                                surface fractions of the more heterogeneous (lower and upper)
                                                layers can result in preferential flow of colloidal particles through
                                                the middle layer, similar to that observed for layered, physically
                                                heterogeneous porous media.
                                                     Because subsurface porous media are physically as well as
                                                geochemically heterogeneous, it is of great interest to investigate
                                                the combined  effect of layered  physical and geochemical
                                                heterogeneity. For the layered physically heterogeneous porous
                                                medium shown in Figure 9, we assumed that the geochemical
                                                heterogeneity is layered distributed as well. Figure 11 a illustrates
                                                the results when the central layer has a larger 1(0.025) compared
                                                to the two side layers (1=0.001). It is interesting to note that for
                                                these conditions the preferential flow path (initially caused by the
                                                physical heterogeneity, Figure  9) disappears due to the
                                                geochemical  heterogeneity. On the other hand, the preferential
                                                flow path is  enhanced when the middle layer has a smaller
                                                1(0.001) than the two side  layers (1=0.025),  as shown  in
                                                Figure 11b.  The results  clearly demonstrate that  layered
                                                geochemical  heterogeneity can significantly alterthe preferential
                                                transport of colloidal particles  caused by heterogeneous flow
                                                field.  Hence,  consideration  of physical or  geochemical
                                                heterogeneity alone in  colloidal transport models may result in
                                                erroneous results.
                                                     Colloid  Transport in Randomly  Heterogeneous Porous
                                                Media.  Colloid transport in  randomly heterogeneous porous
                                                media is investigated in this section. The numerical investigation
                                                is carried out fora point injection (atx=0.5m, z=0.5m) with a pulse
                                                duration of 0.1 day. Results are presented as snapshots  of colloid
                                                concentration in the porous medium at f=0.5 day.
                                                     Freeze  [1975] pointed  out that hydraulic conductivity
                                                variations in aquifers are typically lognormally distributed with a
                                                standard deviation (in log base 10 units) ranging from 0.2 to 2.0.
                                                Since  then,  several  field  measurements confirmed  this
                                                observation. For instance, it was reported that the mean hydraulic
                                                conductivity ACforthe Borden site is 0.0072 cm/s with the variance
                                                of InK being 0.29 [Sudicky,  1986]. Hufschmied [1986]  reported
                                                that the mean value of K\s 0.36 cm/s for the Aefligen site with a
   1.00


   0.90

   0.80


   0.70

   0.60


   0.50


   0.40

   0.30

   0.20


   0.10

   0.00
             1.00
0.50
0.00
    0.00
0.50
1.00
1.50
2.00
2.50
3.00
Figure 9.    Effect of layered physical heterogeneity of porous media on colloid transport. The hydraulic conductivity of the central
            layer is K=100m d"1. The hydraulic conductivities of the upper and lower layers are K = 50 m d"1. The geochemical
            heterogeneity for all the layers is the same, 1=0.01; ccL/aT=5. The xandzaxes show distance (m). The bar graph shows
            the colloid concentration in the porous  medium normalized to the  influent colloid concentration at x = 0.
                                                          28

-------
                       1.00
                                                                                                            (a)
                         0.00
                                      0.50
                                                  1.00
                                                              1.50
                                                                          2.00
                                                                                      2.50
                                                                                                  3.00
               1.00

               QSO

               0«J
               H5G

               .0.40

               0.30

               050

               O'.IO'

               0.00
                                                                                                            (b)
o.oa
   o.oo
                                      0.50
                                                  1.00
                                                               1.50
                                                                           2.00
                                                                                       2.50
                                                                                                    3.00
Figure 10.   Effect of the ratio of aL to aTon the preferential flow of colloids caused by layered physical heterogeneity of porous media:
             (a) 0^/0,.= 1; (b) aL/ccT= 10. The xandy axes show distance (m). The bar graph shows the colloid concentration in the
             porous medium normalized to the influent colloid concentration at x = 0.
                                                                                                             (a)
                                                                                       2.50
                                                                                                   3.00
                l.oo

                0.98
                          1.00-
                         0.50
                         0.00-
                            0.00
                                                                                                             (b)
                                        0.50
                                                                           2.00
                                                                                       2.50
                                                                                                   3.00
Figure 11.    Effect of layered geochemical and physical heterogeneity of porous media on the preferential flow of colloids: (a) central
             layer, K= 100m d~1,  1= 0.025; upper and lower layers contain K= 50 md'1, 1=0.001; aL/aT=5; (b) central layer, K= 100
             md"1, 1=0.001; upper and lower layers contain K= 50 md~\ 1=0.025; aL/aT=5. The x and y axes show distance (m). The
             bar graph shows the colloid concentration in the porous medium normalized to the influent colloid concentration atx = 0.
                                                            29

-------
    1.00

    0.90

    0.80

   10.70

    0.60

    0.50

    0.40

    0.30

    0.20

    0.10

    0.00
0.00
   0.00
0.50
1.00
1.50
2.00
2.50
3.00
Figure 12.   The preferential flow of colloids caused by the layered geochemical heterogeneity in a physically homogeneous porous
             medium. The geochemical heterogeneity of the central layer is X=0.001; the geochemical heterogeneity of the upper and
             lower layers is  1=0.025. Hydraulic conductivity K= 100 m d~1 for each layer; ocL/aT=5. The x and zaxes show distance
             (m). The bar graph shows the colloid concentration in the porous medium normalized to the influent colloid concentration
             atx=0.
variance of InK of 2.15.  The horizontal (x, y directions) and
vertical (zdirection) correlation scales were reported to be 0.29,
2.8, and 0.12m, respectively, fortheBorden site [Sudicky, 1986];
0.26, 5.1, and 0.26 m, respectively, for the Cape Cod site [Hess,
1989]; and 0.031, 3.0,  and 0.91  m, respectively, for the Twin
Lakes site [Moltyaner, 1986].
     In our study, the mean value of Kwas set at 0.116 cm s~1
(100 m d~1), the variance of InK was set at 0.24 or 2.4, and the
correlation scale of the vertical porous medium was set as 0.5 m.
The  random field of hydraulic  conductivity  was  generated
numerically as outlined earlier in this paper, and was incorporated
into the  MCB code of colloid transport. Similar variance values
were used to generate the random field of the particle deposition
rates. The mean values of the deposition rate coefficients were
set at 6.5X10'3  m d~1 for the favorable surface fraction and at
6.5x10"6 m d~1 for the unfavorable  surface fraction. Note that the
latter corresponds to 0^=10~3.
     Realizations of the random fields of hydraulic conductivity
with two different variance values of lnK(0.24 and 2.4) are shown
in  Figure 13. Figure  14 presents the corresponding hydraulic
head distributions and  colloid  concentration  profiles  in the
randomly physically heterogeneous porous media; the results for
physically homogeneous  porous  media are presented as well.
Compared to the physically homogeneous case, random fields of
InK result  in obvious  irregular  hydraulic head distributions
(Figures 14b,c). The irregularity ofthe hydraulic head distributions
increases with  the  variance of  InK A similar  trend can be
observed in the colloid concentration profiles. When the variance
of \nK is small,  the colloid concentration profile (Figure 15b) is
only slightly different than the homogeneous case. However, a
very irregular shape ofthe concentration profile (Figure 15c)can
be seen when the variance of InK is large. The  results clearly
demonstrate that  a  random physical heterogeneity of porous
media results in a random behavior of colloid transport as well.
     Because  of lack of  field  measurements on  random
geochemical heterogeneity of  subsurface porous  media, we
                                                   conducted a preliminary numerical investigation on the sensitivity
                                                   ofthe colloid transport behavior to the parameters characterizing
                                                   the geochemical heterogeneity. Results indicated that the mean
                                                   value ofthe geochemical heterogeneity has to be large enough
                                                   (X=0.01)to show the effect of its spatial distribution on the colloid
                                                   concentration profiles. When the  mean value of X is as small as
                                                   0.001, which may be a reasonable value for sandy aquifers with
                                                   negligible  iron oxyhydroxide  coatings, the  distribution  of A,
                                                   apparently does not affect the colloid transport behavior, even
                                                   when A, is assumed to have a lognormal distribution with a rather
                                                   large variance. Therefore, a mean value of X= 0.01 was chosen
                                                   to carry out the rest ofthe numerical investigation. This value is
                                                   quite reasonable forthe geochemical heterogeneity of subsurface
                                                   geological  formations [Heron et  al., 1994;  Kretzschmar et al.,
                                                   1995; Coston et al., 1995]. It was also found that a  lognormally
                                                   distributed field of A, with a relatively small variance does not have
                                                   an observable effect  on the  particle concentration profiles;
                                                   therefore, a variance of InX as large as 2.0 was chosen.
                                                        RealizationsoftherandomfieldsofXareshown in Figure 16.
                                                   Fora normal distribution, the standard deviation was chosen as
                                                   large as 0.005. Of the simulated X values, about 2.5%  are
                                                   negative; these  negative values were truncated from  the
                                                   distribution. Figure 16a shows that the value of A. is  mostly
                                                   distributed from 0.0 to 0.02. For a lognormal distribution, the
                                                   variance of InA. was set as large as 2.0. We truncated about 2 %t
                                                   values which are larger than 1.0. The value of Ovaries mainly
                                                   between 0.001 to  0.2, and some values of A, even reach 1.0
                                                   (Figure 16b).
                                                       The colloid  concentration  profiles for the  two different
                                                   geochemical heterogeneity fields are compared with the case of
                                                   a constant A, in Figure 17. The normally distributed random field
                                                   of geochemical heterogeneity apparently  does not affect the
                                                   colloid concentration profiles (Figure 17b).  A lognormal  field of
                                                   geochemical heterogeneity with  a  large variance (Figure 17c)
                                                   developed only a slight irregularity in the concentration  profile.
                                                   These  results  suggest that the effect of a random field of
                                                          30

-------
            •3600
            •3300
            •3000
            •2700
            •2400
            •2100
            •1800
            •1500
                        1.O
0.5-
                        O.O

                                                                                 (a)
                                                                                                        (b)
                                                                         2.0
                                                             2.5
3.0
                                                            x(m)
Figure 13.  The realizations of random hydraulic conductivity fields. The scale bar on the left represents hydraulic conductivity (m d~1)
            (a) E(K) = 100 m d'1, Var(lnK) = 0.24; (b) E(K) = 100 m d'1, Var(lnK) = 2.4. The xand z axes show distance (m).
                   13.SSQ

                   19-9S5

                   T3.3BB

                   9.975
                                     0.50      1.00       1.50       2.00       2.50       3.00
                                                                                       (b)
                            0.00       0.50      1.00       1.50       2.00      2.50       3.00
                                                                                       (c)
                           0.00       0.50       1.00       1.50      2.00       2.50       3.00
Figure 14.   Thehydraulicheaddistributionsinhomogeneousorrandomlyphysicallyheterogeneousporousmedia.(a)E(K) = 100 md'1,
             Var(lnK) = 0.0; the porous medium is physically homogeneous; (b) E (K) = 100 m cr1, Var(lnK) = 0.24; (c) E(K) = 100 m
             d~1, Var(lnK) = 2.4. The xand zaxes show distance (m).
                                                           31

-------
   "0.50
                              (a)
                                                                                              (b)
                                                          (c)

Figure 15.   The colloid concentration profiles in a homogeneous or randomly physically heterogeneous porous medium for a point
             injection at (0.5 m, 0.5 m)with a duration of 0.1  dfora snapshot taken at 0.5 d: (a) E(K) = 100 md'1, Var(lnK) = 0.0; the
             porous medium is physically homogeneous; (b) E(K) = 100 m d'1, Var(lnK) = 0.24; (c) E(K) = 100 m d'1, Var(lnK) = 2.4.
             The x and z axes show distance (m).
0.95
0.90
0.85 1 00-
0.80
0.75
0.70
0.65
0.60 °-5°-
0.55
0.50
0.45
0-40 n m
i i j_ i i









       0.0'
         0.00
                    0.50
                               1.00
                                         1.50
                                                    2.00
                                                              2.50
                                                                         3.00
                               1.00        1.50        2.00        2.50
                                                                         3.00
                                                                               0.04 n
                                                                               0.03 --

                                                                               0.03
                                                                               0.02

                                                                               0.02 ,
                                                                               0.01

                                                                               0.01
                                                                               0.00 -'

                                                                                  1  37  73  109145181217253289325
(a)
                                                                               LOOEfOO -
                                                                                1.00E-01 -:
                                                                                1.00E-02
                                                                                1.00E-03
                                                                                1.00E-04
                                                                                                                       (b)
Figure 16.   The realizations and value distributions of random geochemical heterogeneity fields. The gray bar values show the range
             of X. '. (a) normal distribution with E(X) = 0.01  and  o(X) = 0.005; (b) lognormal distribution with E(A.) = 0.01, and
             Var(lnX) = 2.0. The xand zaxes show distance (m).
                                                           32

-------
                                0*1-
  omu  o.oo      0.50
                      1.00      1.50      2.00      2.50     3.00
                                                                 0.00	
                                                                ,o  0.00
                                                          (c)
Figure 17.  The effect of random geochemical heterogeneity on colloid transport for a snapshot at t = 0.5 d after a point release at
            (0.5 m, 0.5 m): (a) colloid concentration profile in a porous medium with a constant geochemical heterogeneity, ^=0.01;
            (b) colloid concentration profile in a normally distributed geochemically heterogeneous porous medium, E(X) = 0.01,
                 = 0.005; (c) colloid concentration profile in a lognormally distributed geochemically heterogeneous porous medium.
geochemical heterogeneity on colloid transport is not as strong
as the effect of random physical heterogeneity. Hence, the mean
value of geochemical heterogeneity is more important than its
distribution in modeling colloid transport in heterogeneous porous
media.
Summary and Conclusions
     The major objectives of this research were to (1) examine
the dependence of colloid transport and mobilization on chemical
perturbations, (2) assess the  relative transport of  mobilized
colloids and the chemicals that caused their mobilization, and (3)
develop a colloid transport model that would begin to describe
these effects. Through the field tests, laboratory experiments,
and model development designed to meet these objectives, we
made significant advances toward testing the major hypothesis
driving this research, that the transport of colloids mobilized in a
contaminant plume will be limited bythe advance ofthe chemical
agent causing colloid mobilization.
     The field tests were conducted in the uncontaminated and
secondary sewage-contaminated zones of the ferric oxyhydroxide-
coated quartz sand aquifer at the U.S. Geological Survey Toxic
Hydrology Research Site  on Cape Cod, Massachusetts. These
experiments examined the dependence of colloid transport and
mobilization on chemical perturbations and assessed the relative
transport of mobilized colloids  and  the chemicals that  caused
their mobilization. The transport of mineral (silica and silica-
coated metal oxide) and biological (viruses) colloids were related
to the surface properties  ofthe colloids and aquifer grains (as
measured by zeta potential). Excellent agreement was found
between  the extent of ferric oxyhydroxide  surface  coverage
measured by electron microprobe and estimated bythe collision
efficiencies for the viruses. Increases in pH were most effective
in mobilizing colloids (both natural  and synthetic) and  viruses
because increases in pH above the  pH^ were most effective in
reversing the charge ofthe ferric oxyhydroxide coatings.  In most
cases, the transport of mobilized colloids was limited by the
advance ofthe colloid-mobilizing agent (e.g., decrease in ionic
strength,  anionic  surfactant concentration, reductant
concentration). A notable exception  occurred  when  pH was
increased in one  field experiment- mobilized colloids appear to
have been transported ahead ofthe hydroxide plume owing to
coating of colloids by natural organic matter. The field research
led to the  development of a new class of tracer colloids, silica-
coated metal oxides. The size of the two types of colloids used in
this research, silica-coated zirconia and titania, was controlled by
varying the precipitation conditions.
     The laboratory experiments examined the dependence of
colloid transport and mobilization  on chemical perturbations
undercontrolled conditions and over a greater range of conditions.
They showed that chemical perturbations that cause increasingly
repulsive conditions produced more extensive and more rapid
colloid release. A series of elevated pH experiments conducted
on individual columns containing oriented, undisturbed sediments
provided excellent data for the  assessment of colloid  release
rates, a task that will  be completed in the future. The laboratory
experiments also showed that an increase of pH to too high a
value produces less repulsion and,  hence, less colloid release.
This result duplicated a phenomenon observed in the field in the
injection well ofthe highest pH injection.
     The  modeling effort aimed at simulating  the  processes
controlling colloid transport in a contaminant plume focused on
the development of a two-dimensional colloid transport model
that considers the geochemical and physical heterogeneity ofthe
porous medium  as well  as the dynamic  aspects  of  particle
deposition. While the modeling effort did not achieve the full
objective of simulating the transport of colloids in a contaminant
plume, majoradvanceswere made. Simulationsofcolloidtransport
in layered heterogeneous porous media indicate that both physical
and geochemical heterogeneities play important roles in colloid
                                                          33

-------
transport. Both types of heterogeneities can cause preferential
flow of colloidal particles. The combination of layered physical
heterogeneity  and layered geochemical  heterogeneity may
enhance or reduce the preferential flow  of  colloids.  Hence,
physical and geochemical heterogeneities should be considered
simultaneously in modeling colloidal  transport in layered
heterogeneous porous media. Overall, the numerical investigation
based on the developed 2-D transport model provides a better
understanding of colloid transport in physically and geochemically
heterogeneous subsurface  porous media. Experimental data of
colloid  transport in different heterogeneous porous  media for
laboratory or natural systems are needed to test the  model
predictions. Since the proposed model  is more sophisticated
than existing models for colloid transport in porous media, the
application of this model will be affected by the availability of the
model parameters. The identification of the model parameters in
this 2-D model is important in applying this model in practice and
will be addressed in future work.

Acknowledgments
    We thank the following people for their help during this
project:  Bob Puls (U.S. EPA, R.S. Kerr Laboratory) for project
guidance. Ron Harvey and  Doug Kent (U.S. Geological Survey)
for discussions on the  project direction.  Denis LeBlanc, Kathy
Hess, and Tim McCobb (U.S. Geological Survey, Massachusetts
District) for access to the field site and collection of sediment and
groundwater samples.  Mike Bonewitz (University of Colorado)
and Jenny Baeseman (University of Wisconsin, Stevens Point)
for field work.  Jon Loveland (University of Colorado) for virus
preparation. Dave  Metge  (U.S. Geological  Survey) and Jon
Larson and Dean  Abadzic (University of Colorado) for data
analysis. John Drexler (University of Colorado) for scanning
electron microscopy. Phil Johnson (Notre Dame University) for
streaming potential analysis.

QUALITY ASSURANCE STATEMENT
    All research  projects  funded  by the  U.S.  Environmental
Protection Agency that make conclusions or recommendations
based on environmentally related measurements are required to
participate  in the Agency  Quality Assurance Program. This
project was conducted  under an approved Quality Assurance
Program Plan and the procedures therein specified were used.
Information on the plan and documentation ofthe quality assurance
activities and results are available from the Principal Investigator.

DISCLAIMER
    The U.S.  Environmental Protection Agency through its
office  of Research and Development  partially funded  and
collaborated in the research described here under Cooperative
Agreement No. CR-824593withtheUniversityofColorado. It has
been subjected to the Agency's peer and administrative  review
and it has been approved for publication  as an EPA document.
Mention oftrade names or commercial products does not constitute
endorsement or recommendation for use.

References
Abdel-Salam, A., and Chrysikopoulos, C. V., (1995) Modeling of
    colloid and colloid-facilitated contaminant transport in a two-
    dimensional fracture  with  spatially  variable  aperture.
    Transport in Porous Media, 20, 197-221.
Adamczyk,  Z.,  Siwek, B., and Zembala,  M., (1992) Reversible
    and  irreversible adsorption of particles on  homogeneous
    surfaces. Colloids Surf. A, 62,  119-130.
Allred, B. and Brown, G.O., (1994)Surfactant-induced reductions
    in soil hydraulic conductivity. Ground Water Monit. Remed.
    174-184.
Amirbahman, A.  and Olson,  T.M., (1993) Transport of humic
    matter-coated hematite in  packed  beds.  Environ.  Sci.
    Technol., 27, 2807-2813.
Ard, R. A. (1997) Natural and artificial colloid mobilization in a
    sewage-contaminated  aquifer: Field  and laboratory
    experiments. M.S. Thesis, University ofColorado at Boulder,
    202 pp.
Atherton, J.G. and Bell, S.S., (1983) Adsorption of viruses on
    magnetic particles-l. Adsorption of MS2 bacteriophage and
    the effect of cations, clay and polyelectrolyte. Water Res.,
    17, 943-948.
Backhus, D.A., Ryan, J.N., Groher, D.M., MacFarlane, J.K., and
    Gschwend, P.M., (1993) Sampling  colloids and  colloid-
    associated contaminants in  ground water. Ground Water,
    31, 466-479.
Bales, R.C., Hinkle, S.R., Kroeger, T.W., Stocking, K. and Gerba,
    C.P.,  (1991)  Bacteriophage adsorption  during  transport
    through  porous media: Chemical  perturbations  and
    reversibility. Environ. Sci. Technol., 25, 2088-2095.
Bales, R.C., Li, S., Maguire, K.M., Yahya, M.T., Gerba, C.P., and
    Harvey, R.W., (1995) Virus and bacteria transport in a sandy
    aquifer, Cape Cod, MA. Ground Water, 33, 653-661.
Barber L.B., II, Thurman,  E.M., Schroeder, M.P. and LeBlanc,
    D.R.,  (1988) Long-term fate  of organic  micropollutants in
    sewage-contaminated groundwater. Environ. Sci. Technol.,
    22, 205-211.
Bellin, A.,  Salandin, P., and Rinaldo, R.,  (1992) Simulation of
    dispersion in heterogeneous  porous formations: Statistics,
    first-order theories, convergence of  computations.  Water
    Resour. Res., 28(9), 2211-2227.
Bixby, R.L. and O'Brien, D.J., (1979) Influence of fulvic acid on
    bacteriophage adsorption and complexation in soil.  Appl.
    Environ. Microbiol., 38, 840-845.
Black, T. C., and Freyberg, D. L, (1987) Stochastic modeling of
    vertically averaged concentration uncertainty in a  perfectly
    stratified aquifer. Water Resour. Res., 23, 997-1004.
Bosma, W. J., and van derZee, S. E. A. T. M., (1993) Transport
    of reacting  solute  in  a one-dimensional,  chemically
    heterogeneous porous medium. Water Resour. Res., 29(1),
    117-131.
Buddemeier, R.W. and Hunt, J.R., (1988) Transport of colloidal
    contaminants in groundwater: Radionuclide migration at the
    Nevada Test Site. Appl. Geochem. 3, 535-548.
Surge, W.D. and Enkiri,  N.K., (1978) Virus adsorption by five
    soils. J. Environ. Qual., 7, 73-76.
Charney, J., Machlowitz,  R.A., and Spicer,  D.S., (1962) The
    adsorption of poliovirus antigen to glass surfaces. Virology,
    18, 495-497.
Chrysikopoulos, C. V., Kitanidis, P. K., and Roberts, P. V., (1990)
    Analysis of one-dimensional solute transport through porous
    media with spatially variable retardation factor. Water Resour.
    Res., 26, 437-446.
Corapcioglu, M. Y., and  S. H. Kim (1995) Modeling facilitated
    contaminant transport by mobile bacteria. Water Resour.
    Res., 31(11), 2639-2647.
                                                         34

-------
Coston, J.A., Fuller, C.C., and Davis, J.A., (1995)Pb2+andZn2+
    adsorption by a natural aluminum- and iron-bearing surface
    coating on an aquifersand. Geochim. Cosmochim. Acta, 59,
    3535-3547.
Dagan, G.  (1984)  Solute transport in heterogeneous porous
    formations.  J. Fluid Mech.,  145, 151-177.
Davis, J.A. (1982) Adsorption of natural dissolved organic matter
    at  the oxide/water interface. Geochim. Cosmochim. Acta,
    46, 2381-2393.
Degueldre,  C., Baeyens, B.,  Goerlich, W., Riga, J., Verbist, J.,
    and  Stadelmann, P., (1989) Colloids in  water from a
    subsurface  fracture in granitic  rock, Grimsel Test  Site,
    Switzerland. Geochim. Cosmochim. Acta, 53, 603-610.
Dick, S.G.,Fuerstenau, D.W., and Healy.T.W., (1971)Adsorption
    of alkyl benzene sulfonate (A. B.S.) surfactants atthe alumina-
    water interface. J. Colloid Interface Sci., 37, 595-602.
Dunnivant, F., Schwarzenbach,  R.P., and Macalady, D., (1992)
    Reduction of substituted nitrobenzenes in aqueous solutions
    containing natural organic  matter. Environ. Sci. Technol.,
    26, 2133-2141.
Elimelech, M., Gregory,  J., Jia, X.,  and Williams, R.,  (1995)
    Particle Deposition & Aggregation, Butterworth-Heinemann
    Ltd.
Elimelech, M., and Song,  L, (1992) Deposition of colloids in
    porous media: Theory and numerical solution. In Transport
    and  Remediation of Subsurface Contaminants: Colloidal,
    Interfacial, and Surfactant Phenomena, D. A. Sabatini and
    R.  C. Knox, Eds.; American Chemical Society Symposium
    Series, 491, 26-39.
Elimelech, M. (1991)  Kinetics of capture of colloidal particles in
    packed beds under  attractive double layer interactions. J.
    Colloid Interface  Sci., 146,  337-352.
Elimelech, M., and O'Melia, C. R., (1990) Kinetics of deposition
    of colloidal particles in porous media. Environ. Sci. Technol.,
    24, 1528-1536.
Farrah, S.R. and Preston, D.R.,  (1991) Adsorption of viruses by
    diatomaceous earth coated  with metallic oxides and metallic
    peroxides. Water Sci. Technol., 24, 235-240.
Freeze, R. A. (1975)  A stochastic-conceptual analysis of one-
    dimensional groundwaterflow in nonuniform homogeneous
    media. Water Resour. Res., 11(5), 725-741.
Fuhs, G.W., Chen, M., Sturman, L.S., and Moore, R.S., (1985)
    Virus adsorption to mineral surfaces is reduced by microbial
    overgrowth  and organic coatings. Microb. Ecol., 11, 25-39.
Garabedian, S. P., Leblanc, D. R., Gelhar, L. W., and Celia, M. A.,
    (1991)  Large-scale natural  gradient tracer test in sand and
    gravel, Cape Cod, Massachusetts, 2,  Analysis of spatial
    moments for a nonreactive  tracer. Water Resour.  Res., 27,
    911-924.
Gelhar, L. W., Welty, C., and Rehfeldt, K. R., (1992) A critical
    review of data on field-scale dispersion in aquifers. Water
    Resour. Res., 28(7), 1955-1974.
Gelhar, L. W., and Axness,  C. L., (1983) Three-dimensional
    stochastic analysis of macrodispersion in aquifers. Water
    Resour. Res., 19(1), 161—180.
Gelhar, L. W., A. L. Gutiahr, and R. L. Naff (1979) Stochastic
    analysis  of macrodispersion  in stratified aquifer. Water
    Resour. Res., 15(6), 1387-1397.

Gerba, C.P. (1984) Applied and theoretical  aspects  of virus
    adsorption to surfaces. Adv. Appl. Microbiol., 30, 133-169.

Gerba, C.P., Goyal, S.M., Cech,  I., and Bogdan, G.F., (1981)
    Quantitative assessmentoftheadsorptive behaviorofviruses
    to soils. Environ. Sci. Technol., 15, 940-944.

Goldenberg,  L.C.,  Magaritz,  M., and  Mandel, S., (1983)
    Experimental  investigation  on irreversible changes  of
    hydraulic conductivity on the seawater-freshwater interface
    in coastal aquifers. Water Resour. Res., 19, 77-85.

Grolimund, D., Borkovec, M., Barmettler,  K., and Sticher, H.,
    (1996) Colloid-facilitated  transport of strongly  sorbing
    contaminants in natural  porous media: A laboratory column
    study. Environ. Sci. Technol., 30(10), 3118-3123.

Harmand, B.,  Rodier, E., Sardin, M.,  and Dodds, J., (1996)
    Transport and  capture of submicron particles in s natural
    sand: short column experiments and a linear model.  Colloids
    and Surfaces A 107, 233-244.

Harvey, R.W. and Garabedian, S.P. ,(1991) Use of colloid filtration
    theory in modeling movement of bacteria  through a
    contaminated sandy aquifer.  Environ. Sci. Technol.,  25,
    178-185.
Harvey,  R.W., George, L.H., Smith, R.L.,  and LeBlanc, D.R.
    (1989) Transport of microspheres and indigenous  bacteria
    through a sandy aquifer: Results  of natural- and forced-
    gradient tracer experiments. Environ. Sci. Technol., 23, 51-
    56.

Harvey, R.W., Kinner, N.E., Bunn, A., MacDonald, D., and Metge,
    D., (1995) Transport behavior of groundwater protozoa and
    protozoan-sized microspheres in sandy aquifer sediments.
    Appl. Environ. Microbiol., 61, 209-217.

Heron, G., Crouzet,  C.A., Bourg, C. M., and Christensen, T. H.,
    (1994) Separation of Fe(ll)  and Fe(lll)  in contaminated
    aquifer sediments using chemical extraction techniques.
    Environ. Sci. Technol., 28,  1698-1705.

Hess, K.H., Wolf,  S.H.  and Celia,  M.A.,  (1992) Large-scale
    natural gradient tracer test in a sand  and gravel  aquifer,
    Cape Cod, Massachusetts 3. Hydraulicconductivity variability
    and calculated macrodispersivities. Water Resour. Res., 28,
    2011-2027.

Hess, K. M. (1989) Use of  borehole flowmeter  to determine
    spatial  heterogeneity  of hydraulic conductivity  and
    macrodispersivity in  a sand and gravel aquifer, Cape Cod,
    Massachusetts, paper presented at NWWA Conference on
    New  Field Techniques  for Qualifying the Physical and
    Chemical properties of Heterogeneous Aquifers Natl. Water
    Well Assoc., Houston, Tex.

Higgo, J. J. W., Williams,   G.  M.,  Harrison,  I., Warwick,  P.,
    Gardiner, M. P., and Longworth G., (1993) Colloid transport
    in a glacial sand aquifer- laboratory and field studies.  Colloids
    Surf. A 73, 179-200.
                                                         35

-------
Hoeskema, R. J., and Kitanidis, P. K., (1984) An application ofthe
    geostatistical  approach  to the inverse problem  in two-
    dimensional groundwater modeling. Water Resour. Res.,
    20(7), 1003-1020.
Hufschmied, P. (1986) Estimation ofthree-dimensional statistically
    anisotropic  hydraulic conductivity field by means of single
    well pumping tests combined with flowmeter measurements.
    Hydrogeologie, 2, 163-174.
Johnson, P.R., Sun, N. and Elimelech,M.,(1996)Colloidtransport
    in geochemically heterogeneous porous media: Modeling
    and measurements. Environ. Sci. Technol., 30, 3284-3293.
Johnson, P. R., and M. Elimelech (1995) Dynamics of colloid
    deposition  in  porous media: Blocking  based on  random
    sequential adsorption. Langmuir, 11, 801-812.
Kent, D.B., Davis J.A., Anderson L.C.D., Rea B.A. and WaiteT.D.
    (1994) Transport of chromium and selenium in the  suboxic
    zone of a shallow aquifer: Influence of redox and adsorption
    reactions. Water  Resour. Res., 30, 1099-1114.
Khilar, K.C. and Fogler,  H.S., (1984) The existence of a critical
    salt concentration for particle release. J. Colloid Interface
    Sci., 101, 214-224.
Kolakowski, J.E. and Matijevic, E., (1979) Particle adhesion and
    removal in model systems Part 1 .-Monodispersed chromium
    hydroxide on glass. J. Chem. Soc. Faraday Trans. I 75, 65-
    78.
Kretzschmar,  R., Robarge, W.P. and Amoozegar,  A., (1995)
    Influence of natural organic matter  on colloid transport
    through saprolite. Water Resour. Res., 31,  435-445.
Lapidus,  L, and  Amundson,  N.R.,  (1952)  Mathematics of
    adsorption in beds. VI, The effect of longitudinal diffusion in
    ion exchange and chromatographic columns. J. Phys. Chem.,
    56,  984-988.
LeBlanc, D. R.  (1984)  Sewage  plume in  a sand and gravel
    aquifer, Cape Cod, Massachusetts. U.S. Geol. Survey Water
    Supply Paper 2218, 28 pp.
LeBlanc, D.R.,  Garabedian,  S.P., Hess,  K.H., Gelhar, L.W.,
    Quadri, R.D.,  Stollenwerk, K.G. and Wood, W.W., (1991)
    Large-scale natural gradient tracer test in sand and gravel,
    Cape Cod, Massachusetts  1.  Experimental design  and
    observed tracer movement. Water Resour. Res., 27, 895-
    910.
Leij, F. J., and Dane, J. H., (1990) Analytical solutions of the one-
    dimensional  advection equation  and  two- or three-
    dimensional dispersion equation. Water Resour. Res., 26(7),
    1475-1482.
Liang, L. and Morgan, J.J. ,(1990) Chemical aspects of iron oxide
    coagulation in water: Laboratory studies and implications for
    natural systems. Aquatic Sciences 52, 32-55.
Lindqvist, R., and  Enfield, C. G., (1992) Cell density and non-
    equilibrium sorption effects on  bacteria  dispersal in
    groundwater microcosms. Microb. Ecol., 24, 25-41.
Liu, H. H., and Molz, F. J., (1997) Comment on "Evidence  for non-
    Gaussian scaling behavior  in heterogeneous sedimentary
    formations" by Scott Painter. Water Resour. Res., 33(4),
    907-908.
Loveland, J.P., Ryan, J.N., Amy, G.L. and Harvey, R.W. (1996)
    The reversibility  of virus attachment to  mineral surfaces.
    Colloids Surf. A 107, 205-221.

Magelky, R.  D. (1998) Colloid deposition and mobilization in a
    sewage-contaminated iron oxide-coated aquifer. M.S. Thesis,
    University of Colorado, Boulder, 235 pp.

Mantoglou,  A., and Wilson,  J.  L., (1982) The turning  bands
    method for simulation of random fields using line generation
    by a spectral method. Water Resour. Res.,  18(5),  1379-
    1394.

McCarthy, J. F.,  and Degueldre, C., (1993)  Sampling  and
    characterization of colloids and particles in groundwater for
    studyingtheirroleincontaminanttransport. In Environmental
    Particles, Environmental Analytical and Physical Chemistry
    Series, Buffle, J., and H. P. van Leeuwen (Eds.),  Lewis
    Publishers.

McCarthy, J. F., andZachara, J. M., (1989) Subsurface transport
    of contaminants.  Environ. Sci. Technol., 23, 496-502.

McDowell-Boyer, L.M. (1992) Chemical mobilization of micron-
    sized particles in  saturated porous media under steady flow
    conditions. Environ. Sci. Technol., 26, 586-593.

Mills A.L., Herman, J.S., Hornberger, G.M. and DeJesus, T.H.
    (1994) Effect of solution ionic strength and iron coatings on
    minerals grains on the sorption of bacterial cells to  quartz
    sand. Appl.  Environ. Microbiol., 60, 3300-3306.

Moltyaner, G. L. (1986) Stochastic versus deterministic: A case
    study. Hydrogeologie, 2, 183-196.

Molz, F. L.,  Guven, O.,  Melville, J. G.,  Crocker, R. D.,  and
    Matteson, K. T., (1986) Performance, analysis, and simulation
    of a two-well tracer test at the Mobile site. Water Resour.
    Res., 22(7), 1031-1037.

Moore, R.S., Taylor,  D.H.,  Reddy, M.M.M. and Sturman,  LS.
    (1982) Adsorption of reovirus by minerals and soils. Appl.
    Environ. Microbiol., 44, 852-859.

Moore, R.S., Taylor, D.H., Sturman, L.S., Reddy, M.M. and Fuhs,
    G.W. (1981) Poliovirus adsorption by 34 minerals and soils.
    Appl. Environ. Microbiol., 42, 963-975.

Mukerjee,  P.  and  Mysels, K.  J.,  (1971) Critical micelle
    concentrations of aqueous  surfactant systems. NSRDS-
    NBS 36. Washington,  DC,  National Bureau of Standards,
    227 pp.

Murray,  J.P. and Parks,  G.A., (1980) Poliovirus adsorption on
    oxide  surfaces.  Correspondence with  the DLVO-Lifshitz
    theory of colloid stability. In Particulates  in Water.
    Characterization, Fate, Effects,  and Removal (Kavanaugh
    M.C. and Leckie J.O.,  Eds.), p.  97. American  Chemical
    Society, Washington, DC.

Nightingale, H.l.andBianchi, W.C., (1977)Ground-waterturbidity
    resulting from artificial recharge. Ground Water, 15,146-152.

Olsen, R.H., Siak, J.-S. and Gray, R.H. (1974) Characteristics of
    PRD1, a plasmid-dependent broad  host  range  DNA
    bacteriophage. J. Virology,  14, 689-699.
                                                         36

-------
Painter, S. (1996) Evidence for non-Gaussian scaling behaviorin
    heterogeneous sedimentary formations. WaterResour. Res.,
    32(5), 1183-1195.
Painter, S. (1997) Reply. WaterResour. Res., 33(4), 909-909.
Parks,  G.A. (1965) The isoelectric points of solid oxides, solid
    hydroxides, and aqueous hydroxo complex systems. Chem.
    Rev., 65, 177-198.
Parks,  G.A. (1967) Aqueous surface chemistry of oxides and
    complex oxide minerals. In Equilibrium Concepts in Natural
    WaterSystems(StummW., Ed.), p. 121. American Chemical
    Society, Washington, DC.
Penrod, S.L., Olson, T.M., and Grant, S.B., (1996) Deposition
    kinetics  of two viruses in packed beds of quartz granular
    media. Langmuir 12, 5576-5587.
Pieper, A.P., Ryan, J.N., Harvey, R.W., Amy, G.L., Illangasekare,
    T.H., and Metge, D.W., (1997) Transport and recovery of
    bacteriophage PRD1 in a sand and gravel aquifer: Effect of
    sewage-derived organic matter. Environ. Sci. Technol., 31,
    1163-1170.
Privman,  V., Frisch, H.  L, Ryde,  N., and Matijevic,  E., (1991)
    Particle adhesion in model system. 13. Theory of multilayer
    deposition.  J.  Chem. Soc. Faraday  Trans., 87(9), 1371-
    1375.
Puls, R.W., Clark, D.A., Bledsoe, B., Powell, R.M. and Paul, C.J.,
    (1992) Metals in ground water: sampling  artifacts  and
    reproducibility. Haz. Waste Haz.  Mat, 9, 149-162.
Puls, R.W. and  Powell, R.M.,  (1992) Transport of inorganic
    colloids  through  natural  aquifer  material:  Implications for
    contaminant transport. Environ. Sci. Technol., 26, 614-621.
Rajagopalan, R. and Tien, C., (1976) Trajectory analysis of deep-
    bed filtration with the sphere-in-cell porous media model.
    AlChE J. 22, 523-533.
Redman,  J.A., Grant, S.B., Olson,  T.M., Hardy, M.E. and Estes,
    M.K., (1997) Filtration of recombinantNorwalk virus particles
    and  bacteriophage MS2 in  quartz sand:  Importance of
    electrostatic interactions. Environ. Sci. Technol., 31, 3378-
    3383.
Roy, S.B. and Dzombak, D.A., (1996) Colloid release and transport
    processes in natural and model porous media. Colloids Surf.
    A 107, 245-262.

Ryan, J.N., Ard, R.A.,  Magelky,  R.D., and Elimelech, M., (in
    preparation) Effect of elevated pH on the mobilization of
    colloids from a ferric oxide-coated quartz sand: Laboratory
    and field experiments. To be submitted to Environmental
    Science and Technology.

Ryan, J.N. and  Elimelech, M., (1996) Colloid  mobilization and
    transport in groundwater. Colloids Surf. A 107, 1-52.

Ryan, J.N., Elimelech, M., Ard, R.A., Harvey, R.W., and Johnson,
    P.R., (1999) Bacteriophage PRD1 and silica colloid transport
    and recovery in an iron oxide-coated sand aquifer. Environ.
    Sci. Technol., 33, 63-73.
Ryan, J.N. and Gschwend, P.M.,  (1990) Colloid mobilization in
    two Atlantic Coastal  Plain aquifers:  Field studies. Water
    Resour. Res., 26, 307-322.
Ryan, J.N. and Gschwend, P.M., (1991) Extraction of iron oxides
    from sediments using reductive dissolution by titanium(lll).
    Clays Clay Min., 39, 509-518.
Ryan, J.N. and  Gschwend,  P.M.,  (1994)  Effects of solution
    chemistry on clay colloid mobilization from an iron oxide-
    coated sand  aquifer. Environ. Sci. Technol., 28, 1717-1726.
Ryan, J.N., Magelky, R.D., and Elimelech, M., (in preparation)
    Using silica-coated metal oxide tracer  colloids of variable
    size in a colloid transport field experiment. To be submitted
    toColloidsandSurfacesA. Physicochemicaland Engineering
    Aspects.
Saiers, J. E., Hornberger, G. M., and  Harvey, C., (1994) Colloidal
    silica transport through  structured, heterogeneous  porous
    media. J. Hydrology, 163, 271-288.
Scholl, M.A. and  Harvey, R.W. (1992) Laboratory investigations
    on the role of sediment surface  and groundwater chemistry
    in transport of bacteria through a contaminated sandy aquifer.
    Environ. Sci. Technol. 26, 1410-1417.
Small, D.A. and  Moore, N.F. (1987) Measurement of surface
    charge of baculovirus polyhedra. Appl. Environ. Microbiol.,
    53, 598-602.
Smith, R.L., Howes, B.L. and Garabedian,  S.P. (1991) In-situ
    measurement of methane oxidation in groundwater using a
    natural gradient tracer test. Appl. Environ. Microbiol.,  57,
    1997-2004.
Sobsey, M.D., Dean, C.H., Knuckles, M.E.,  and Wagner, R.A.,
    (1980) Interactions and survival of enteric viruses in  soil
    materials. Appl. Environ. Microbiol., 40, 92-101.
Song, L., and M. Elimelech (1994) Transient deposition of colloidal
    particles  in  heterogeneous  porous  media.  J. Colloid and
    Interface  Sci., 167, 301-313.
Song, L., Johnson, P.R., and Elimelech, M., (1994) Kinetics of
    colloid deposition  onto heterogeneously charged surfaces
    in porous media. Environ. Sci. Technol., 28, 1164-1171.
Sudicky, E. A. (1986) A natural gradient  experiment on solute
    transport in a sand aquifer: Spatial variability of hydraulic
    conductivity  and its role in the  dispersion process. Water
    Resour. Res., 22(13), 2069-2082.
Sun, N. (1998) Colloid transport in physically and geochemically
    heterogeneous subsurface  porous  media, Ph.D.  Thesis,
    University of California at Los Angeles.
Sun, N., Elimelech,  M., Sun, N.-Z.,  and Ryan J.N. (submitted)
    Colloid transport  in  physically and  geochemically
    heterogeneous subsurface porous media: 1.    A   two
    dimensional model. Submitted to WaterResources Research.
Stumm, W., and Morgan, J. J., (1995) Aquatic Chemistry, 3rd Ed.
    Wiley-lnterscience: New York.
Sun, Ne-Zheng (1995) Mathematical Modeling of Groundwater
    Pollution, Springer-Verlag, New York.
Sun, Ne-Zheng, and Yeh, W. W.-G., (1983) A proposed upstream
    weight numerical method for simulating pollutanttransport in
    groundwater. Water Resour. Res., 19(6), 1489-1500, 1983.
Tobiason, J. E. (1989) Chemical effects on the deposition of non-
    Brownian particles. Colloids Surf., 39, 53-77, 1989.
                                                         37

-------
Tompson,  A.  F.  B., Ababou, R.,  and Gelhar,  L. W., (1989)
    Implementation  of the three  dimensional turning bands
    random field  generator. Water Resour. Res., 25(10), 2227-
    2243,  1989.
Walter, D. A., Rea, B. A., Stollenwerk, K. G., andSavoie, J. (1996)
    Geochemical and  hydrologic controls  on phosphorus
    transport in a sewage-contaminated sand and gravel aquifer
    near Ashumet Pond, Cape Cod, Massachusetts. U.S. Geol.
    Survey Water-Supply Paper 2463, 89 pp.
Yao K.-M., Habibian, M.T. and O'Melia, C.R. (1971) Water and
    wastewater treatment filtration: Concepts and applications.
    Environ. Sci.  Technol. 5, 1105-1112.
                                                       38

-------