You are on page 1of 7

Environ. Sci. Technol.

1997, 31, 1191-1197

Soil Chemistry Effects and Flow flows is thus of great importance. For electromigration, Hicks
and Tondorf (10) showed how products of electrode reactions
Prediction in Electroremediation of could halt the removal of heavy metal contaminants by
affecting the metal speciation. By controlling the cathode’s
Soil product, OH-, they achieved removals of over 95%. For
electroosmosis, Shapiro and Probstein (6, 8) showed that in
some cases convective flow ceased before high removal
JOHN M. DZENITIS*
percentages were reached. They found that by using a basic
Department of Mechanical Engineering, Massachusetts purge solution to limit the anode’s product, H+, they could
Institute of Technology, Cambridge, Massachusetts 02139 promote flow toward the cathode. The specific mechanism
for electroosmotic flow cessation has not been as conclusively
identified as in the electromigration case, largely because of
the complexity of the multispecies electrochemical transport.
This work addresses processes occurring during the The most physically realistic model of electroremediation
removal of contaminants from soils using electric fields. transport was introduced by Shapiro et al. (5) and extended
Laboratory experiments and mathematical modeling are used and generalized by Jacobs et al. (14) and Jacobs and Probstein
to study the changes in the flows of ions and pore liquid (15). Despite the detail of this model, the electroosmotic
during the process; these flows are directly related to the flow velocity is based on the measured flow rate, so the model
removal of charged and uncharged contaminants by elec- is not predictive in terms of convective velocity. In particular,
tromigration and electroosmosis, respectively. Soil properties the causes of varying flow rate and flow cessation, so important
are explored by electrophoresis measurements, acid/ to contaminant removal by electroosmosis, cannot be
determined. Eykholt was the first to include dependencies
base titrations, and elemental analyses of pore solutions,
required to model changing electroosmotic flow during
then incorporated into an electrochemical transport model, electroremediation (16-18). His model did show varying flow
and compared to electroremediation experiments. It is found rate, and he was able to predict a change in the direction of
that a soil chemistry model involving cation exchange and electroosmotic flow when acid was added to the cathode
aluminum chemistry must be included to describe the reservoir. However, there was not quantitative agreement
process accurately. Varying electroosmotic flow is suc- between his experiments and numerical simulations. In this
cessfully predicted, but only until the development of a paper, we develop the first quantitatively accurate predictions
low ionic strength region in the medium. The insight gained of varying charge and mass flow during electroremediation.
allows the mechanisms of electroosmotic flow reversal Insights into the mechanisms of flow cessation are uncovered
and cessation to be identified. As importantly, this investigation in the process.
finds the low ionic strength region to be an undesirable
but likely occurrence with or without significant effects from Experimental Section
soil chemistry and shows how controlling the system Electroremediation Experiments. The apparatus used for
chemistry makes the electroremediation technique more the one-dimensional electroremediation experiments is
robust in practice. shown schematically in Figure 1 and described in detail in
ref 19. The soil mixture with length ≈150 mm was contained
in a clear PVC tube with an inner diameter of 54 mm. The
ends of the soil were held by filter paper against a stainless
Introduction steel screen that acted as a mechanical support and voltage
Electroremediation is an innovative method for removing probe. Electrode reservoirs on either side of the soil contained
contaminants from soil using in situ, low power, dc electric carbon fiber electrodes (Fiber Materials Inc., Biddeford, MA)
fields. Like other in situ methods such as bioremediation, across which the voltage was applied. The cathode reservoir
vapor extraction, and soil flushing, electroremediation has (165 mL) was connected to a tank on a scale, and the anode
advantages in avoiding high costs and human health risks of reservoir was part of a gravity-fed recirculation system (total
excavation. Additionally, electroremediation is well-suited volume of 4500 mL) using a return pump with wetted surfaces
to heavy metal contaminants, unlike bioremediation and of polypropylene. This large volume gave the anode reservoir
vapor extraction, and it is applicable to contaminants in a high chemical capacitance, useful for controlling the system
heterogeneous and low-permeability soils, unlike soil flushing. chemistry as described below. The pressure at the ends of
Charged contaminants such as heavy metals in solution are the cell was balanced by adjusting the feed tank height, so
primarily moved by electromigration, and uncharged con- all of the measured mass flow resulted from electroosmosis
taminants such as soluble organic molecules can be moved (2). The dc power supply was adjusted throughout the
with the bulk liquid in the presence of charged soil surfaces experiment to maintain 15 V across the 150 mm soil length.
by electroosmosis (1, 2). Once the contaminants reach the Measurements of current, applied voltage, effluent mass,
electrode reservoirs, the solutions can be easily pumped out applied soil matrix stress, and local voltage and pressure in
and treated. the soil were made with a digital data acquisition system.
Laboratory experiments (3-11) and limited field work (12, The soil used was an acidic form of a nearly pure kaolin
13) have proven that it is possible to achieve nearly complete clay (Albion Sperse 100, Albion Kaolin Co., Hephzibah, GA).
removal of contaminants using electric fields. However, these A barely-liquid mixture was made by gradually stirring the
studies have also shown that the approach can fail when flow dry clay into a 10 mM NaCl solution until a solid:liquid mass
of charge (current) or flow of mass (convection) are not ratio of 1:1 was reached. The loading piston was used to
maintained. Determining the causes of decrease in these gradually consolidate this mixture in the test cell to assure
a tight seal with the walls.
* Present address: Monsanto Company U4E, 800 North Lindbergh To investigate the effect of chemical changes on the
Boulevard, Saint Louis, MO 63167; telephone: 314-694-8696; fax: 314- process, two electroremediation experiments are presented
694-1531; e-mail address: jmdzen@ccmail.monsanto.com. here. In the first experiment (untreated), the main electrode

S0013-936X(96)00707-9 CCC: $14.00  1997 American Chemical Society VOL. 31, NO. 4, 1997 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 1191
FIGURE 1. Apparatus used for electroremediation experiments. Measurements of current i, voltage distribution V, effluent mass m, and
soil matrix stress s are indicated.

products, H+ and OH-, were not controlled. As the experi-


ment progressed, the pH at the anode reservoir dropped to
2, and the pH at the cathode reservoir increased to 11. In the
second experiment (base addition), the anode product H+
was replaced with Na+ by periodically adding concentrated
NaOH to the reservoir, which kept the anode pH above 9.
This approach is similar to that taken by Shapiro and Probstein
(6, 8).
The electroosmotic flow rate through the medium can be
characterized with an average electroosmotic permeability
ka ) u/Ea (m2 V-1 s-1), where u is the volumetric flow per unit
area per unit time or average interstitial velocity (m s-1) and
Ea is the applied electric field (V m-1). In one dimension, Ea
is the voltage drop across the medium divided by its length.
The use of ka for flow of mass emphasizes that the flow results
solely from the electric field and forms a useful analogue
with average conductivity and flow of charge later. The initial
conditions are uniform, and if the charged double layer at
the solid/liquid interface is thin as compared to the inter- FIGURE 2. Electroosmotic flow rate in terms of average electroos-
particle distance, the electroosmotic permeability is given by motic permeability ka ) u/Ea for experiments with and without base
addition at the anode. At 7 days, the untreated experiment has
ζ displaced approximately 1 pore volume and shows flow cessation,
ka ≈ - at t ) 0 (1) while the base addition experiment has displaced over 2.2 pore
µτ2
volumes and is increasing.
where  is the permittivity of the liquid (6.93 × 10-10 C V-1
m-1 in 298 K water), µ is the viscosity (kg m-1 s-1), τ is the of the pore solution modified by the tortuosity factor:
dimensionless porous medium tortuosity, and ζ is the
F2
ζ-potential (V), which is identified with the electric potential
at the soil/liquid interface (2). The tortuosity is a constant
σa ≈ ∑z ν c
τ2
2
j j j at t ) 0 (2)
(g1) introduced to account for the indirect path through the
porous medium. As the experiment progresses, the average where F is Faraday’s constant (96 487 C mol-1), zj is the charge
electroosmotic permeability involves a spatial average because number of the species j, νj is the mobility (mol s kg-1), and
conditions in the medium become non-uniform (5, 8). cj is the concentration (mol m-3 or mM). When the chemical
Equation 1 no longer applies, but the experimental definition composition changes, σa becomes a complicated spatial
ka ) u/Ea still holds. The experimental electroosmotic average involving varying concentrations, diffusion, local
permeability is plotted in Figure 2. The initial behavior of electric field effects, and the surface conductivity (5). The
the two experiments was identical, but the flows diverged experimental definition σa ) i/Ea still holds at these later times,
after 2 days. The experiment with base addition showed and these results are shown in Figure 3. The initial behavior
increasing flow rate while the untreated experiment’s flow of the two experiments was again identical up to 2 days and
began to cease at 7 days. At this point, the experiment with then diverged as the conductivity of the base addition
base addition had displaced over 2.2 pore volumes while the experiment began to rise dramatically at about 4 days.
untreated experiment had displaced only 1. Soil Medium Transport Properties. To determine why
The charge flow analogue to the mass flow above is mass and current flow decreased in the first case and dropped
interstitial current density i, which is movement of charge but later increased in the second case, a detailed numerical
per unit area per unit time (A m-2). As above, the porous model was used to track the multispecies transport including
medium’s condition can be represented by a single value, in electromigration, electroosmotic convection, diffusion, and
this case the average electrical conductivity σa ) i/Ea (S m-1). chemical reactions. In this work, we used a one-dimensional
Initially the concentrations are uniform and surface con- version of the model introduced by Shapiro et al. (5) and
ductivity can be neglected, so σa is simply the conductivity extended by Jacobs et al. (14) and Jacobs and Probstein (15).

1192 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 31, NO. 4, 1997


FIGURE 3. Charge flow rate in terms of average electrical
conductivity σa ) i/Ea for experiments with and without base addition FIGURE 4. ζ-potential measurements and empirical fit for varying
at the anode. Both experiments show an initial drop in conductivity, pH and at three Na+ concentrations. The model assumes a point
but the base addition experiment’s conductivity begins to rise of zero net charge at pH 6 and linear dependence on pH and the
dramatically at ≈4 days. logarithm of ionic strength.

The deviations and details are not reproduced here. The form can be physically justified, but the number of data points
required solvent properties (density, viscosity, permittivity) used here is really insufficient for the curve fit. This is certainly
were taken to be those of pure water, and the solute properties an area that could bear further work. The pH dependence
(diffusion coefficient, mobility, chemical equilibrium coef- is similar to the measurements of Lorentz (22), which had
ficients) were taken to be those of the infinitely dilute species less variation in ionic strength. Eykholt and Daniel (17) used
in water. The porous medium properties (porosity, tortuosity, Lorenz’s data to include ζ-potential dependence on pH.
hydraulic permeability, surface conductivity, surface poten- Soil Chemical Behavior. The chemical properties of the
tial, chemical behavior) depend on the specific soil and local soil medium were the final part of the system to be
conditions in the medium, so these properties were deter- characterized. Chemical behavior is important because it
mined separately. The results are summarized below, and determines the species that are present, the electric field
more details can be found in ref 19. distribution (via the conductivity distribution), and the soil
A porosity of 0.54 was measured from consolidation tests, surface charge; in other words, chemical interactions deter-
and a tortuosity of 1.65 was calculated from eq 2 using an mine what is present and how it moves. Despite its
initial conductivity measurement. The hydraulic permeability importance, the issue of soil chemistry has been avoided in
was measured to be 4 × 10-16 m2 at the initial conditions, and electroremediation work because of its complexity. Here,
a surface conductivity of 10-3 S m-1 measured by Shapiro (6) soil chemistry was tackled in a manner similar to that used
for the same clay was used. for the surface potential above. The key, again, is to realize
The ζ-potential of the surface seen in eq 1 is a property that electromigration separates the background electrolyte
depending on complex physicochemical interactions (20). ions and replaces them with electrode products, so that acid/
Given the scale of the transport problem and the complex base (HCl/NaOH) titration is the appropriate way to explore
composition of soils, empirical data and major simplifications the soil’s chemical behavior.
are required. Microelectrophoresis measurements of the Alkalinity is the concentration of strong base minus the
kaolin clay’s ζ-potential were made with a Zeta-Meter 3.0+ concentration of strong acid in a solution. For a system with
(Zeta-Meter Inc., Long Island, NY) in hydrosols of various only H+, OH-, Na+, and Cl-, the alkalinity is given by Alk )
composition. Since pH, ionic strength I ) 1/2∑z2j cj, and [Na+] - [Cl-]. In this case, the concentrations of all species
exchangeable cation concentration are key parameters in are known when alkalinity and ionic strength are given if
determining ζ-potential, the solutions were designed to electroneutrality, ∑zjcj ) 0, and water equilibrium, [H+][OH-]
explore these dependencies. The pH was varied with HCl ) Kw, are assumed to always hold (23). The titrations were
and NaOH because NaCl was used as a background electrolyte performed with initial alkalinity -100 e Alki e 100 mM using
in the electroremediation experiments; it can be shown that NaOH for positive Alki and HCl for negative values. Instead
the electroremediation process effectively forms this acid and of trying to completely cover the two-dimensional alkalinity/
base by separating the salt’s ions (19). The initial ionic ionic strength space, two extremes were taken in the titrations:
strength and cation concentration were varied by adding (a) Constant initial ionic strength with NaCl added as
NaCl to each of the solutions so they had one of three Na+ necessary to make Ii ) 100 mM.
concentrations: 0.1, 10, or 200 mM. Measurements at 1 and (b) Minimum initial ionic strength with no NaCl added so
6 days showed that there was generally little change in Ii ≈ |Alki|.
ζ-potential over this period. In making the soil mixtures, a compromise was struck
The measured ζ-potentials are shown in Figure 4 together between reproducing the electroremediation test conditions
with a simple empirical fit of the data. The pH had a great and having a workable mixture. A solid:liquid mass ratio of
effect on the clay, resulting in ζ-potentials from +10 mV to 1.5:1 was selected for the initial pH experiments because the
-45 mV. Sodium concentration had an effect only for pH > resulting slurry could still be stirred but was close to the
6. In this range, the ionic strength is equal to the sodium concentration in the electroremediation experiments (2.2:1).
concentration, so there was a decrease in ζ-potential mag- The concentration effect was incorporated later.
nitude for higher ionic strengths. This behavior is consistent The pH electrode measurements after 24 h are shown in
with more extensive work on clays and metal oxides (20, 21). Figure 5 versus initial liquid alkalinity. The pH curve for a
The empirical fit shown in Figure 4 was based on depend- solution without weak acid or base is also plotted for reference.
encies seen in these works; we assumed a point of zero net The relative flatness of the experimental curves represents a
charge at pH 6 and linear dependence on pH and log I. The buffering resistance to pH change. The clay shows some

VOL. 31, NO. 4, 1997 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 1193


FIGURE 5. Experimental titration of kaolin clay showing buffering
to both acid and base relative to a solution with no weak acid/base
behavior. There is little difference between the constant and varying
ionic strength paths.

buffering of acid and more pronounced buffering of base,


which is expected because this particular clay is in acidic
(H+) form. The acidic nature of the clay is seen as well as in
fact that with no added alkalinity, the initial pH is less than
7.
Ionic strength had little effect on the titration curve. This
is an unexpected result in some ways because surface
chemistry models usually include ionic strength dependence,
and simple ion exchange models would predict a dependence
on sodium concentration. On the other hand, the relative
insensitivity to background electrolyte compared to H+ is
common and is the same sort of behavior seen for the surface
potential. There is some deviation between the two curves FIGURE 6. Clay chemical effect in titrations in terms of measured
for small positive alkalinities in Figure 5, and those differences change in pore liquid concentration. Significant consumption of
are consistent with higher Na+ concentration displacing more sodium (a) and release of aluminum (b) are seen.
H+ from the clay surface. The differences are small on the
overall scale, however, and the dependence on ionic strength Changes in Al, Ca, Fe, K, and Mg are only seen as releases
will be neglected from now on. since none of these species were present in the liquid initially.
The greatest effect was from Al release at both negative
Since the clay was seen to have a significant chemical
and positive alkalinities. In addition, Ca and Mg were
effect, another series of titrations was performed together
released in similar amounts in negative alkalinity. There was
with elemental analyses of the resulting pore solutions. A
no change in Fe or K concentration. All of these measured
slightly lower solid:liquid mass ratio (1.2:1) was used to make
clay element responses are consistent with the pH results;
it easier to obtain liquid for the analysis, and only the constant
they are changes that buffer the system pH to both base and
ionic strength titration was performed. The mixtures were
acid addition.
prepared as before, and 24 h later pore liquid samples were
It is clear that the clay’s reactive concentration range (≈100
separated from the mixtures by centrifuging through tubes
mM) is significant on the scale of the initial electrolyte
with internal membranes (Ultrafree CL 0.45 µm, Millipore
concentration (10 mM). What is not clear is how the soil
Inc., Bedford, MA). The pore solutions were analyzed for Na,
affects electroremediation and how the soil behavior should
Cl, Al, Ca, Fe, K, and Mg using inductively coupled plasma
be incorporated in the transport model for predictive
(ICP) spectroscopy in a Perkin-Elmer Plasma 40 (Norwalk, purposes. The simplest means of incorporating soil chemistry
CT). Since Cl cannot be detected with this device, its would be to use the mineral’s equilibrium equations, but it
concentration was determined indirectly by AgCl precipita- can be shown that the measured species and concentrations
tion. In retrospect, since kaolinite is an aluminosilicate, silicon do not correspond to kaolinite equilibrium. This is not
should have been added to the elements analyzed. The clay surprising given natural impurities and the long time scale
behavior can be modeled without silicon, but better results of mineral equilibrium. Creating artificial equilibrium models
might have been possible had it been included. can be successful, however, because the time scale of simple
The results of the elemental analyses are shown as changes surface reactions [minutes (23)] is short as compared to the
in pore liquid concentration multiplied by magnitude of the time scale of electroremediation. Three different models of
ion charge in Figure 6. Presenting the data in this way weights the soil behavior were constructed for use in numerical
the elements according to their charge contribution. This is simulations of the electroremediation experiments. These
not strictly true for aluminum, however, since Al(OH)4- will models are briefly introduced below, then included in the
form at high pH. The major trend seen is consumption of electroremediation model, and compared to the data in the
Na+ in the positive alkalinity range. This mechanism can section Applications of Models to Experiments.
explain the buffering to base (NaOH) seen in Figure 5, as will When soil chemistry is ignored here, the species present
be discussed below. The changes in Cl- and in Na+ near zero are H+, OH-, Na+, and Cl-; the chemical reactions are water
alkalinity may not be significant; in this range, [NaCl] ≈ 100 electrolysis at the electrodes and water equilibrium through-
mM and the accuracy of the analysis are probably not better out the liquid. Soil chemistry is included by introducing
than 10% for Cl- and 5% for Na+. additional species and chemical equilibrium reactions to the

1194 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 31, NO. 4, 1997


chemical system. The simplest model of the measured
buffering behavior (Figure 5) is an amphiprotic acid/base
site XH capable of accepting or donating a proton. The solid
then acts as a weak acid and a weak base, buffering the
medium to additions of base and acid through the reactions

XH + Na+ + OH- ) X- + Na+ + H2O


(3)
XH + H+ + Cl- ) XH2+ + Cl-

A partitioning function fit to the measured pH response was


used. This was taken to be a function of pH only because the
next likely dependence, ionic strength, was seen to have little
effect in the titration.
The acid/base site model is designed to match the
experimental pH response, but it ignores the ion consumption
and release seen in Figure 6. Ion exchange behavior is well
known in the field of soil chemistry (see, e.g, ref 24) and can
be used to reproduce the cation consumption and release
seen in the elemental analysis results. Since Na+ was the
major participant, a sodium ion exchange site is used in the
second soil chemistry model, giving buffering reactions:

XH + Na+ + OH- ) XNa + H2O


(4)
XNa + H+ + Cl- ) XH + Na+ + Cl-

Since H+ and Na+ are the only cations in this model, Na+
release serves as a substitute for the actual release of Al3+,
Ca2+, and Mg2+. Again, the partitioning function was made
to fit the data in Figure 5.
The third soil chemistry model reproduces the experi-
mental aluminum release by including solid aluminum
hydroxide Al(OH)3(s) as a species. This is largely insoluble FIGURE 7. Average conductivity for experiments and (a) NaCl
at zero alkalinity, but dissolves when sufficient amounts of simulations (ignoring soil chemistry) and (b) ion exchange/aluminum
acid or base is added, which approximates the aluminum hydroxide soil model simulations with experimental mass flow as
release seen in Figure 6. The initial amount of Al(OH)3(s) an input. The poor agreement in panel a and the success in panel
was based on the maximum dissolved concentration, and b shows that the medium chemistry is not properly predicted without
literature values (23) were used for the solubility product and including soil effects.
equilibrium constants of the dominant dissolved species, Al3+
and Al(OH)4-. As above, sodium ion exchange was used to Better results were obtained using the acid/base and cation
give the Na+ consumption behavior and match the experi- exchange soil models, but close agreement in both conduc-
mental titration pH. tivity and voltage distribution required the ion exchange/
aluminum hydroxide model (Figure 7b).
Applications of Models to Experiments In the next set of simulations, the single “free” input to
Simulations of electroremediation transport were run without the runs abovesthe electroosmotic flow velocityswas no
soil chemistry and with soil chemistry models based on acid/ longer specified. Instead, the empirical surface potential
base, ion exchange, and ion exchange/aluminum hydroxide model shown in Figure 4 was used in determining the local
reactions. The first set of simulations focused on soil contributions to the electroosmotic flow. The inputs were
chemistry effects, and the experimental electroosmotic all determined from independent experiments, and the
permeability (Figure 2) was used as an input. All other evolution in time was determined solely by the numerical
parameters were either determined from independent ex- model. Figure 8 shows the flow rate as average electroosmotic
periments or calculated in the simulation. Even using the
permeability for experiments and simulations with and
experimental electroosmotic permeability, the transport
without soil chemistry. As expected from the conductivity
problem is quite complex: the concentration and electric
results, the case ignoring soil chemistry (Figure 8a) could not
potential distributions must be found, and the migration,
predict the electroosmotic flow accurately. The ion exchange/
diffusion, and chemical reactions of all species must be
aluminum hydroxide model gave much better results (Figure
determined as they progress in time and space. The soil
8b), properly predicting initially negative flow (towards the
surface species (XH, XNa, X-, XH2+) and solid aluminum
anode), early increase, and subsequent plateaus. Once the
hydroxide are immobile, but their local concentrations change
flow models diverged significantly from the experiments, the
as the transport shifts the mobile species in solution.
species’ distributions became incorrect, and the simulated
Comparisons between the simulations and experiments can
be made in terms of the medium’s local and average system became unstable. The instability of the system and
conductivity. the divergence of both experimental and modeled flow around
Figure 7 shows the average conductivity for experiments 1.7 days will be discussed below.
and simulations with experimental mass flow as an input. Because the empirical surface potential model (Figure 4)
The average conductivity involves a spatial integration of the was a weak link in the analysis, another simulation based on
species’ concentrations throughout the medium (related to Lorenz’s data (22, 17) was run. This showed a faster initial
eq 2), and the poor agreement in Figure 7a shows that the increase and higher plateau than seen in Figure 8b, but gave
composition of the pore solution was not properly predicted. a similar overall behavior.

VOL. 31, NO. 4, 1997 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 1195


unlikely that the pH in the high field region is exactly pH0.
Although the model here did not predict flow cessation, the
reason for its occurrence is indicated. The fundamental
physicochemical effect is the development of large Debye
length (thickness of the charged liquid layer adjacent to the
charged clay solid) in the low ionic strength region formed
near the cathode. An analysis with simplified geometry (25)
shows that the electroosmotic permeability (e.g., eq 1) should
be multiplied by a factor that decreases as ionic strength
decreases; for the smaller ionic strengths in these simulations
(I < 0.1 mM), the attenuation factor can be less than 0.4,
which would explain discrepancies such as those seen in
Figure 8b. Eykholt (16) includes this effect, but does not
emphasize the results. Simplified corrections involve as-
sumptions that are not quantitatively applicable to our
situation, but the general concept applies. Low ionic strength
is related to low conductivity and the high field region, which
means that most of the electrical effort is being applied right
where it is least effective. Also, there is a positive feedback
effect where the low ionic strength causes higher electric field
strength, which accelerates the deionization. The resulting
attenuation factor could reduce the flow by orders of
magnitude. One important conclusion is that significant flow
reductions or cessation can occur even in soils that do not
show zero ζ-potential behavior.
The low ionic strength region explains two types of
deviations seen at about 1.7 days in Figure 8b: (1) the
experimental cases’ flow rates diverge because the base
addition begins to have an effect on the low ionic strength
region, and (2) the simulations diverge from the experiments
because the model does not take into account the electroos-
motic attenuation. The absence of this attenuation also leads
to the instability seen in the simulations.
There is also a mechanical effect that may play a role in
some observations: the high field region creates low pressure
and large pressure gradients, which may not be supportable
FIGURE 8. Electroosmotic flow rate in terms of electroosmotic in some apparatuses. Low pressures can cause external leaks,
permeability ka ) u/Ea for experiments and (a) NaCl simulations and large pressure gradients can cause “short-circuiting” of
(ignoring soil chemistry) and (b) ion exchange/aluminum hydroxide the liquid flow along the walls of a test cell. The apparatus
soil model simulations with no free inputs. A soil chemistry model and in-cell consolidation used here prevented these effects
is necessary for quantitative predictions, and even then experiment
from being a problem, but they are likely to occur in other
and model diverge at ≈1.7 days.
setups. Low pressures and high pressure gradients in the
field may cause uneven velocity distributions and channeling,
Discussion which could lead to expending energy to move liquid that is
Quantitative prediction of electric and electroosmotic flow is not contaminated. Low pressure could also cause evolution
important in improving understanding of the processes of dissolved gases, breaking the ionic conductance in the
occurring during electroremediation. One phenomenon that region.
can now be better understood is electroosmotic flow initiation. Electromigration is an important transport mechanism in
Negative initial flow is not surprising given that (1) the electroremediation whether electroosmosis is the primary
electroosmotic permeability can be negative (ζ > 0) for acidic mechanism for contaminant removal or not. The changes
pH and (2) clays that give an acidic initial pH are often used in conductivity seen here are driven mainly by electromi-
in laboratory work. The more interesting question is why the gration transport, with shifting of the distributions from
flow becomes positive. The answer is that the local electric convective movement. The early drop in overall conductivity
field becomes large precisely where the local electroosmotic seen in Figure 7 is a result of the local conductivity changes
permeability takes on significantly positive values, increasing discussed above; the drop occurs before there is much
the flow toward the cathode. This is predicted even in the convective displacement. The low ionic strength region can
absence of soil chemistry because the NaOH solutions formed also form when there is no flow of the pore liquid at all, with
near the cathode have lower electrical conductivity (hence undesirable effects for electromigration transport itself. First,
higher field) than the HCl solutions created near the anode. the region is often associated with a jump in pH that can
A related but more pronounced effect is seen when the proper change the sign of the charge of the dominant heavy metal
soil chemistry model is included; the soil reaction near the species, leading to focusing of the contaminant within the
cathode (eq 4a) removes charged species and markedly medium (1, 10). Second, low conductivity means that the
decreases the local conductivity. This makes a small region overall movement of ions by electromigration is slow, so the
near the cathode (10-20% of the total medium length) develop remediation process would take a great deal of time.
a large electric field and provide virtually all of the positive Controlling the system chemistry is the key to avoiding
pumping power. the formation of the low ionic strength region. The region
Another phenomenon, at least as important as electroos- results from reactions that eliminate charge-carrying ions
motic flow initiation, is the electroosmotic flow cessation (e.g., H+ and OH-), in interactions either with each other or
seen in this and other experimental work. The most prevalent with the soil surface. By substituting less reactive species at
qualitative explanation of flow cessation involves the soil’s one or both electrodes, this eliminating reaction is avoided.
point of zero net charge (pH0, where ζ ) 0). However, it is The resulting electric field distribution is more even, and flow

1196 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 31, NO. 4, 1997


of charge and mass are maintained. If soil chemistry is (6) Shapiro, A. P. Ph.D. Dissertation, Massachusetts Institute of
negligible, the added ions can have an immediate and strong Technology, Cambridge, MA, 1990.
effect, as is seen in Figure 7a. If the soil participates (7) Hamed, J. T.; Acar, Y. B.; Gale, R. J. ASCE J. Geotech. Eng. 1991,
117(2), 241-271.
significantly in reactions with the background electrolyte or
(8) Shapiro, A. P.; Probstein, R. F. Environ. Sci. Technol. 1993, 27(2),
the added chemicals, the effect of the chemical addition may 283-291.
be delayed, as shown in Figure 7b. (9) Acar, Y. B.; Alshawabkeh, A. N. Environ. Sci. Technol. 1993, 27(13),
Some soil types (e.g., sandy ones) may have little chemical 2638-2647.
capacitance compared to the background electrolyte and (10) Hicks, R. E.; Tondorf, S. Environ. Sci. Technol. 1994, 28(12), 2203-
could be ignored in the system’s chemical model; others (e.g., 2210.
those rich in humates, clays with smaller particle size) may (11) Acar, Y. B.; Gale, R. J.; Alshawabkeh, A. N.; Marks, R. E.; Puppala,
have significantly higher reactive concentrations than those S.; Bricka, M.; Parker, R. J. Hazard. Mater. 1995, 40, 117-137.
observed here. The specific soil model constructed here will (12) Lageman, R. Environ. Sci. Technol. 1993, 27(13), 2648-2650.
not be applicable for most natural soils, but the framework (13) Athmer, C. J.; Ho, S. V. Presented at the ACS Symposium Emerging
and procedure used will be useful and should be included as Technologies in Hazardous Waste Management VII, Atlanta, GA,
Sep 18, 1995.
part of site characterization and process design. It is
(14) Jacobs, R. A.; Sengun, M. Z.; Hicks, R. E.; Probstein, R. F. J. Environ.
important to first understand how electroremediation could Sci. Health 1994, A29(9), 1933-1955.
change the chemical composition throughout the medium (15) Jacobs, R. A.; Probstein, R. F. AIChE J. 1996, 42(6), 1685-1696.
and then explore the soil response to the expected changes (16) Eykholt, G. R. Ph.D. Dissertation, The University of Texas at
in batch experiments. A simplified soil chemistry model can Austin, 1992.
then be constructed and used to further refine the chemical (17) Eykholt, G. R.; Daniel, D. E. ASCE J. Geotech. Eng. 1994, 120(5),
control and process design. 797-815.
(18) Eykholt, G. R.; Daniel, D. E. ASCE J. Geotech. Eng. 1996, 122(3),
Acknowledgments 252-254.
(19) Dzenitis, J. M. Ph.D. Dissertation, Massachusetts Institute of
The author thanks R. F. Probstein and R. E. Hicks for their Technology, Cambridge, MA, 1996.
guidance during the time this work was performed. Financial (20) Hunter, R. J. Zeta Potential in Colloid Science; Academic Press:
support was provided in part by the Office of Science and London, 1981.
Technology within the U.S. Department of Energy’s Office of (21) Dzombak, D. A.; Morel, F. M. M. Surface Complexation Modeling:
Environmental Restoration and Waste Management under Hydrous Ferric Oxide; Wiley: New York, 1990.
the Contaminant Plumes Containment and Remediation (22) Lorenz, P. B. Clays Clay Miner. 1969, 17, 223-231.
Focus Area, by the U.S. Environmental Protection Agency (23) Morel, F. M. M.; Hering, J. G. Principles and Applications of
Northeast Hazardous Substance Research Center at New Aquatic Chemistry; Wiley: New York, 1993.
Jersey Institute of Technology, and by MIT’s John Hennessy (24) Bolt, G. H.; Bruggenwert, M. G. M.; Kamphorst, A. Adsorption
Fellowship for Environmental Studies. of cations in soil. In Soil Chemistry A. Basic Elements; Develop-
ments in Soil Science Vol. 5A; Bolt, G. H., Bruggenwert, M. G.
M., Eds.; Elsevier: Amsterdam, 1978.
Literature Cited (25) Rice, C. L.; Whitehead, R. J. Phys. Chem. 1965, 69(11), 4017-
(1) Probstein, R. F.; Hicks, R. E. Science 1993, 260, 498-503. 4024.
(2) Probstein, R. F. Physicochemical Hydrodynamics: An Introduction,
2nd ed.; Wiley: New York, 1994.
(3) Runnells, D. D.; Larson, J. L. Ground Water Monit. Rev. 1986, Received for review August 16, 1996. Revised manuscript
6(3), 85-91. received November 22, 1996. Accepted December 6, 1996.X
(4) Renaud, P. C.; Probstein, R. F. PCH, PhysicoChem. Hydrodyn.
1987, 9(1/2), 345-360.
ES960707E
(5) Shapiro, A. P.; Renaud, P. C.; Probstein, R. F. PCH, PhysicoChem.
Hydrodyn. 1989, 11(5/6), 785-802. X Abstract published in Advance ACS Abstracts, February 15, 1997.

VOL. 31, NO. 4, 1997 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 1197

You might also like