Water Research 37 (2003) 1818–1830

Nitrate removal in zero-valent iron packed columns

Paul Westerhoff a,*, Jennifer Jamesb
a
Department of Civil and Environmental Engineering, Arizona State University, P.O. Box 5306, Tempe, AZ 85287-5306, USA
b
Sewerage System Section, Los Angeles County Water Authority, 920 South Alameda Street, Compton, CA 90221, USA

Abstract

Nitrate removal by laboratory and field continuous-flow zero-valent iron (Fe0) packed bed columns was evaluated
for different influent water qualities (pH, dissolved oxygen (DO), nitrate concentration) and several months of
operation (600–1500 bed volumes (BVs)). In contrast to previous batch experiments with Fe0 where nitrate was
stoichiometrically converted to ammonium, only 70% of the applied nitrogen was recovered as nitrate, ammonium, or
nitrite (o0.1 mg/L) during shorter-term column tests (2–20 BVs) and less than 25% of the applied nitrogen was
recovered during longer-term field testing (500–1000 BVs) at elevated nitrate levels (B25 mg N/L). Nitrate removal was
accompanied by a pH increase, DO decrease, and soluble iron increase. During longer-term operation (500–1500 BVs)
iron and calcium precipitates were observed, by SEM and EDX analyses, to form in the packed columns. Precipitation
led to cementation and reduction in permeability for the Fe0/sand media in the packed column. Different abiotic and
microbial-mediated mechanisms may be involved during shorter- and longer-term operation of Fe0 systems and the role
of iron precipitates should be further evaluated.
r 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Nitrate; Zero valent iron; Corrosion

1. Introduction Both mechanisms rely upon the oxidation of Fe0 and

Zero-valent iron (Fe0) has been proposed for use for
in situ groundwater reactive barriers for oxo-anions
(e.g., nitrate, chromate) and chlorinated organics [1–7].
Laboratory studies have demonstrated the feasibility for
electro-chemical reduction to occur, and hence the
viability of Fe0 as a treatment technology. Mechanisms
for electro-chemical reduction are still unresolved in
many cases, and may involve direct electron exchange
between a contaminant (e.g., nitrate) and Fe0 that is
reduced to ferrous or ferric ion, or production of
hydrogen gas (H2) which subsequent exchanges elec-
trons with a contaminant:
Fe0 -Fe2þ þ2e ;
Fe0 þ 2Hþ -Fe2þ þH2 :
*Corresponding author. Tel.: +1-480-965-2885; fax: +1-
480-965-0557.
E-mail address: p.westerhoff@asu.edu (P. Westerhoff).
production of ferric or ferrous ions. The fate of these
iron by-products on sustainability of Fe0 processes has
received relatively little attention, yet may have the
potential of affecting the long-term viability of Fe0
treatment systems.
Previous work has demonstrated the ability for zero-
valent metals (e.g., Fe0) to reduce nitrate in batch
experiments, with nearly stoichiometric recovery of
nitrogen as ammonium [8–13]. Nitrate (NO 3 ) is a target
compound of concern since it occurs in ground and
surface waters as a consequence primarily of agricultural
fertilization or wastewater discharges, and has regula-
tory health limits in the US (MCL=10 mg N/L) and EU
(25 mg NO3/L). The primary health effect is actually
ð1Þ related to nitrite (NO2 ), which forms upon reduction of
nitrate in the gastrointestinal tract, and can lead to
ð2Þ methoglobinemia, a potential fatal acute disease result-
ing from oxygen deprivation [14]. Nitrate released to the
environment can also increase eutrophication levels in
rivers or lakes [15]. Current biological and physico-
chemical treatment systems are capable of removing

0043-1354/03/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0043-1354(02)00539-0
 P. Westerhoff, J. James / Water Research 37 (2003) 1818–1830 1819

nitrate, but may be costly or difficult to operate in many for changes in surface morphology and mineralogy
situations [16]. Fe0 offers the potential as a treatment (energy dispersive X-ray (EDX) analysis). The results of
process that could operate in situ to remediate ground- this study suggested that increased attention be focused
water (e.g., permeable barrier walls) or as above-ground on cementation of Fe0 systems, fate of by-products (e.g.,
packed-bed treatment processes. nitrate by-products and iron by-products), and changes
Since nitrate reduction by Fe0 is a corrosive process, in nitrate-reduction mechanisms (e.g., electrochemical
the formation of passivating scales on the Fe0 over time reduction, sorption, microbial mediate reactions).
may limit its long-term reduction potential. Few long-
term studies are available that report on these phenom-
ena. Farrell et al. [4] investigated the long-term 2. Methods and materials
performance of Fe0 for TCE reduction in packed
columns fed with different ionic mixtures. TCE reaction 2.1. Laboratory Fe0 column experiments
rates in the presence of calcium sulfate or calcium
chloride were unaffected over a 2-year period, while Laboratory column test experiments were conducted
TCE reaction rates decreased in the presence of calcium with a native groundwater (Tempe, AZ) spiked with
nitrate. The nitrate contributed to increased iron surface nitrate at a rehabilitated wellsite. Table 1 summarizes
passivity. Phillips et al. [17] reported the effects of the average water quality conditions for the native
mineral precipitates and corrosion for an in situ reactive groundwater. Water quality was typical of regional
barrier, after 15 months of operation, which was used to wells, characterized by DO concentrations of 4–6 mg/L.
remove uranium. Scanning electron microscopy (SEM) The well serves as a hydrologic linkage between two
of iron cores from different locations showed spatial different aquifer units. Localized pumping affected
variability. Cementation of Fe0 was occurring due to the water levels in the aquifer units, which have different
formation of iron hydroxides, which lead to localized nitrate levels. Hence, during testing, localized pumping
decreases in permeability. Aragonite crystals (CaCO3), affected influent nitrate levels and water quality.
iron sulfides (FeS), and hexagonal-shaped goethite (b- Lab column tests were conducted using plexi-glass
FeOOH) minerals were observed. In situ treatment columns (600mL; 5 cm diameter; 30 cm height). The Fe0
processes have typically been studied under low velocity columns were packed in 2–5 cm lifts by adding pre-
conditions that simulate groundwater movement, and weighed dry Fe0 (1636 g at 6.32 g/cc of FE5; 2271 g at
limited work is available on above-ground higher-rate 7.89 g/cc of FE7) to distilled water. A lower support
packed-bed treatment systems. Reduction in permeabil- layer of 3 cm of clean sand was packed in the column
ity and formation of passivating layers that may limit first, and 2 cm was placed above the Fe0 material. The
mass transport of nitrate to the reactive surface are the final column porosity was approximately 40%. Previous
two primary concerns for long-term sustainability of the batch experiments screened several types of Fe0 for
in situ Fe0 systems. nitrate removal capability [18], and two iron types were
The objectives of the study were to evaluate both
nitrate removal in and by-product release from Fe0
packed beds under laboratory and field conditions with Table 1
model and natural waters. Two hypotheses were Average water chemistry of groundwater used for study (SRP
evaluated. First, nitrate removals and by-products are Well ID 23.0E-2.9N; [44])
constant during shorter- and longer-term Fe0 packed Parameter Parameter value
column operations, and supported by batch experimen-
tal results. Second, nitrate corrodes Fe0 and leads to Total dissolved solids (ppm) 790
soluble iron production and cementation of packed Major anions
columns. Two sources of Fe0 were tested in laboratory Chloride (ppm) 200
Sulfate (ppm) 90
continuous flow columns, based upon nitrate removals
Fluoride (ppm) 0.2
observed in previous batch experiments that showed Nitrate (ppm) 5.1
different levels of performance for nitrate removal [18]. Major cations
Several influent water quality conditions (nitrate con- Sodium (ppm) 205
centration, pH, dissolved oxygen (DO)) were evaluated Potassium (ppm) 5.4
in the laboratory, prior to operating packed columns Calcium (ppm) 62
extended durations (hundreds to thousands of bed Magnesium (ppm) 25
volumes). Field experiments were conducted to evaluate pH 7.7–8.0
pilot-scale operational performance (nitrate removal, Alkalinity (ppm as CaCO3) 304
hydraulic sustainability, by-product production) of Hardness (ppm as CaCO3) 256
Dissolved oxygen (mg/L) 4–6
above-ground treatment systems. Field columns were
Dissolved organic carbon (mg/L) o0.2
sacrificed and the packing material examined by SEM
1820 P. Westerhoff, J. James / Water Research 37 (2003) 1818–1830
selected for column experiments. Masterbuilders Inc.
scrap iron filings sieved to 30  200 mesh (acronym:
FE5) had a surface area of 1.2 m2/g, and Baker Iron
Chips (0.5–5 mm) had a surface area of 0.016 m2/g
(acronym: FE7). Surface areas were determined by a
nitrogen sorption BET test (Science and Technology
Laboratory, Miami, FL). Prior to use, four bed volumes
of 0.01 N sulfuric acid was passed through the column to
remove any iron oxides. Several bed volumes of model
groundwater were then passed through the column prior
to collecting reported data. A peristaltic pump (Cole
Palmer Inc.) applied the test solution in upflow mode
through the column. Upflow mode limited the potential
for gas entrainment. Different empty bed contact times
(i.e., EBCT equaled column volume divided by flowrate)
were achieved by varying the pump flowrate. Influent
and effluent samples were collected.
During preliminary tests with the shorter-term
laboratory columns operated for up to 20 bed volumes
(BVs), it was observed that after 2 BVs the performance
(C/Co for nitrate) was constant. This was not surprising
since: (1) the columns behaved nearly as PFRs (see
tracer tests below), and (2) the EBCT was 50% longer
than the actual residence time. Therefore, columns were
operated for 2 BVs prior to collecting samples in the
effluent stream, which were then collected for the
following 3 BVs. Variable initial water qualities were
tested (nitrate concentration, DO, pH). Sodium nitrate,
sodium hydroxide, and sulfuric acid were added as
necessary. For low DO tests the sample water was
sparged with nitrogen gas (99.995% purity). For
dissolved chemical analysis the samples were filtered
(0.2 mm Gelman Nylon), and for nitrate/nitrite analysis
passed through prefilters to remove residual dissolved
iron (Dionex Onguard-H). Average concentrations and
standard deviations were calculated from the replicate
samples. Longer-term laboratory experiments were
conducted using the same Fe0 columns, but were
operated continuously under a single water quality
condition for several weeks.
Bromide tracer tests were conducted during operation
of the laboratory columns (FE5 and FE7) to evaluate
hydraulic performance for a liquid flow rate that
achieved an EBCT of 3 h. In general, the bromide
breakthrough curves were sharp and indicative of a
plug-flow reaction (PFR) with minimal lateral or axial
dispersion. The calculated hydraulic residence time
based upon analysis of the F-curve was 133 min for
FE5 (dispersion number (D/UL)=0.043; N ¼ 12
CSTRs in series) and 123 min for FE7 (dispersion
number (D/UL)=0.040; N ¼ 13 CSTRs in series) [19].
2.2. Field Fe0 packed-bed treatment system
The field columns were packed with FE5 in a similar
manner as the laboratory columns, but were larger in
size (4L; 7.5 cm diameter; 91 cm height) and 8 cm of
support sand at the bottom. Initially, groundwater
(Table 1) was pumped (Grundfow Stainless Steel
10S03-6, 1/3 HP, 0.28 m3/s) up through a 0.4 cm
galvanized pipe, and flow split between the field column
and waste line. The first column ran (Run 1) under
pressure from the submersible pump in the well at
ambient nitrate levels (B5 mg N/L). Later in the study
(Run 2) the column was replaced with a new packed-bed
column and a peristaltic pump provided flow through
the column, and a second peristaltic pump was used to
continuously spike the column influent with nitrate to a
concentration between 5 and 15 mg N/L. The bottom
half of the packed column during Run 2 was a mixture
of 50% sand and 50% FE5, which was intended to
improve sustainability of permeability, whereas Run 1
had 100% FE5 throughout the column. Sodium nitrate
(1 M) dissolved in groundwater was pumped from a
storage container and blended in a static mixer with flow
pumped (peristaltic pump) from the well.
Over the course of approximately 12 months, two
columns packed with FE5 were operated in the field.
Operational problems occurred frequently during the
first run (Run 1: October 1999–March 2000), so the
column was removed and a second column was operated
for several months (Run 2: May 2000–August 2000).
The average EBCT of the field columns was 1.1 h (Run
1) and 1.5 h (Run 2). The most significant operational
problem encountered with the packed-bed treatment
system was a decrease in permeability over time. Effluent
flowrates decreased during both runs. Run 2 had fewer
operational problems than Run 1. Run 1 was not
continuous due to operational problems such as air
entrainment, electrical power outage at the site, and
cracks in reactor. Using the field log of dates when the
column was in continuous operation the cumulative
number of BVs treated was calculated.
2.3. Analytical methods
Reagent grade chemicals were used, and analytical
standard methods (SM) followed [20]. Nitrate, nitrite,
and bromide were measured by ion chromatography
(Dionex CX-120) following SM 4110. Ammonium was
measured using an ISE (Hach Model 50250 electrode
and Corning 340 m) following SM4500-NH3-F-Phenate
Method. Ammonium, nitrite, and nitrate all reported as
mg N/L.
Total iron was measured by atomic adsorption
(Perkin Elmer 3110) following SM 3111B, after sample
acidification. After syringe filtration, a colorimetric
method (SM3500-Fe-D) was used to determine ferrous
iron content (Hach DR2000). DO (SM4500-0 G) was
used (Corning meter 340 and probe 97-08). Solution pH
was measured (Beckman F11 meter and 139841
electrode). Micrograph images of the iron were obtained
 P. Westerhoff, J. James / Water Research 37 (2003) 1818–1830 1821

by SEM. An SEM model JSM-840 (manufactured by
Motorola) equipped with EDX elemental composition
analyzer was used to analyze iron particle surfaces. The
X-ray micro-analyzer detected elements of atomic
number from 11 to 57 (i.e., sodium, Na, to lanthanum,
La,). The upper limit of the detected elements depended
on the maximum acceleration potential (40 keV) for the
SEM model JSM-840. The acceleration potential used
during this investigation was 15 keV. The iron samples
were mounted and fixed on aluminum cylinder stubs of
12-mm diameter and 10-mm height using self-adhesive
carbon conductive tabs.
3. Results
3.1. Shorter-term laboratory column operation
factor of 20 for FE5. This implies that nitrate removal in
FE7 was more dependent upon initial nitrate concentra-
tion than FE5.
DO concentrations in the influent groundwater
(spiked with nitrate) equilibrated with the atmosphere
were between 8.0 and 8.5 mg/L during the experiments.
The effluent DO from the Fe0 columns was typically
between 1 and 3 mg/L. Data in Table 3 present the effect
of initial DO on nitrate removal. Surprisingly, lower
initial DO resulted in 50% less reduction in nitrate. It
had been considered that oxygen reduction would be
necessary prior to an onset in nitrate reduction.
However, the results suggest just the opposite.
During all tests, pH levels increased across the Fe0
packed columns. Table 4 summarizes the effects of

3.1.1. Effect of initial water quality parameters 100

Ambient groundwater (4.6 mg N/L, pH=8.0) spiked

with sodium nitrate was passed through packed bed
columns for several BVs, and the average nitrate 80
removal (1-C/Co) was calculated. Lower influent nitrate
Percentage Nitrate Removed

concentrations and lengthening the EBCT resulted in

higher percentage removal of influent nitrate (Fig. 1). 60
Nitrate removal ranged from 30% to 90% for the FE5
packed column. Normalized mass nitrate removals
(Table 2) were calculated from mass balances on nitrate, 40
flowrate through the column, and volume of the reactor
(600 mL). Calculated nitrate removal rates (mg NO3-N/
4.7 mg-N/L
L-reactor/h) increased with higher initial nitrate con- 13.5 mg-N/L
20
centrations, and decreased slightly with longer EBCTs. 28.0 mg-N/L
Shorter-term column experiments with FE7 had lower 64.5 mg-N/L
29 mg-N/L (pH 7.1)
percentage nitrate removals than FE5 (e.g., Table 3).
Calculated nitrate removal rates (mg NO3-N/L-reactor/ 0
h) for FE7 were comparable with FE5 data at the 0 1 2 3 4

highest initial nitrate concentration (Table 2). However, EBCT (hours)

the nitrate removal rates were 90 times less for the lowest Fig. 1. Percentage nitrate removed as a function of initial
(4.1 mg N/L) compared against the highest (65 mg N/L) nitrate concentration and EBCT in laboratory columns (Iron
initial nitrate concentration for FE7. In contrast, the type=FE5; initial pHB8; error bars represent one standard
difference in calculated nitrate removal rates were a deviation (n ¼ 3)).

Table 2
Normalized nitrate removal rates under varying nitrate (4.7–64.5 mg-N/L) in short-term laboratory experiments (initial pH=8.0)

Fe0 source EBCT (h) Nitrate removal rate (mg NO3-N/L-reactor/h)

4.7 mg-N/L 13.5 mg-N/L 28 mg-N/L 29 mg-N/La 64.5 mg-N/L

FE5 1 2.8 7.4 12.7 11.8 18.6

2 2.1 5.8 10.2 7.8 10.3
3 1.4 4.1 7.6 6.3 10.9

FE7 1 0.08 1.7 3.5 — 13.6

2 0.14 0.82 2.3 — 11.8
3 0.18 1.1 3.5 — 10.1
a
Initial pH=7.0.
1822 P. Westerhoff, J. James / Water Research 37 (2003) 1818–1830

Table 3 35
Effect of initial dissolved oxygen (DO) concentration on nitrate
removal from groundwater (initial nitrate B25 mg-N/L;
pH=8.3–8.5) 30

Iron EBCT Percentage nitrate removal

Ammonia Producted (mg-N/L)

source (h) 25
[DO]0B8.8 ppm [DO]0B0.6 ppm

FE5 1 4571% 1470.3% 20

2 7371 3571
3 8172 3975
FE7 1 1273 4.670.3 15
2 1770.4 1574
3 4074 1174 10

Table 4
Effect of initial pH on nitrate removal (initial nitrate B20 mg-
N/L; EBCT=1 h)
Iron Percentage nitrate removed pH
source (%)
Initial
FE5 4174 7.09
4571 8.23
2275 8.4
FE7 2473 7.46
1273 8.33
370.5 8.5
initial pH and the rise in pH for select experiments,
under the same initial nitrate concentration. Generally,
lower initial pH had larger percentage removals of
nitrate.
3.1.2. Ammonium and iron by-products
Previous batch tests indicated nearly 100% conver-
sion of nitrate to ammonium [12,13,18]. Fig. 2 presents
the yield of ammonium versus the loss of nitrate from all
short-term lab column tests. Relatively uniform conver-
sions of nitrate to ammonium were observed in all the
experiments. Approximately 70% (FE5: 77%; FE7:
65%) of the added nitrate was recovered as ammonium,
based upon slopes of regression lines through the two
sets of data. The incomplete mass balance could be due
to any combination of the following: (1) analytical
standard error, (2) production of nitrogen gases, and (3)
sorption of nitrate/ammonium ion. In a few experi-
ments, when water quality was changed between runs
(i.e., washing with Super-Q water) desorption of nitrate
was observed.
Soluble iron exiting the Fe0 packed columns was
measured and found to be highly variable. The iron
concentration in the influent water was 0.08 mg/L.
Ferrous iron was low compared to the total soluble
iron, and less than 0.01 mg/L in 53% of the measured
samples. In the remaining samples, the ferrous iron
FE7
5 FE5
0
0 5 10 15 20 25 30 35
Nitrate Removed (mg-N/L)
Final
Fig. 2. Summary of nitrate-N removed and ammonium-N
8.25 produced from short-term laboratory column experiments with
8.81 FE5 and FE7 iron types (amended ambient groundwater).
8.68
8.04
8.26 accounted for only 10% of the total soluble iron. The
8.40 total iron concentration in the FE5 effluent averaged
2.3 mg/L (range: 0–15 mg/L) and 8.5 mg/L (range
0–27 mg/L) for FE7. In general, higher nitrate mass
removals correlated to increased iron release and lower
measured DO samples contained higher percentages of
ferrous iron.
3.2. Longer-term laboratory column operation
Two separate packed bed columns (FE5 and FE7)
were continuously operated in parallel for 40 days
(650 BVs) in the lab with groundwater spiked with
nitrate (B25 mg N/L). Data for nitrate and ammonium
in the effluent from the columns are plotted in Fig. 3
(FE5) and Fig. 4 (FE7), respectively. The percentage of
nitrate reduced across the column was variable ranging
from o10% to 85%. In general, nitrate removal
improved during the first 500 BVs. At around 550 BVs,
an unknown event caused poor nitrate removals in both
columns, but may have been related to refreshing of the
influent water, which occurred approximately every 60–
70 BVs. The FE5 column appeared to demonstrate an
improvement in nitrate removal after 600 BVs.
While effluent nitrate concentrations were variable,
effluent ammonium concentrations were relatively con-
stant, accounting for approximately 20% of the influent
nitrate. The sum of nitrate plus ammonium was always
less than the influent nitrogen load. Nitrite was always
o0.1 mg/L. On average only 50% of the nitrate
removed was recovered as ammonium, and was lower
 P. Westerhoff, J. James / Water Research 37 (2003) 1818–1830 1823

35 8
Influent NO3
Effluent NO3
FE5
30 Effluent NH4 7 FE7
Effluent NO3+NH4

6
25
Concentration (mg-N/L)

Effluent Iron (mg/L)

5
20

4
15

3
10

2
5

1
0
0 100 200 300 400 500 600 700 0
BV Treated 0 100 200 300 400 500 600 700
BV treated
Fig. 3. Nitrate-N reduction across a laboratory column with
FE5 during extended operation, and ammonium-N production Fig. 5. Summary of soluble iron in the effluent of laboratory
plus the sum of ammonium-N and nitrate-N in the column columns with FE5 or FE7 during extended operation (ambient
effluent (ambient groundwater with nitrate spiked to B25 mg groundwater with nitrate spiked to B25 mg N/L; initial iron
N/L). B0.08 mg/L).

35
Influent NO3
ammonium ion (NH+ 4 ) sorbed to freshly precipitated
Effluent NO3
30 Effluent NH4 iron oxides and established a pseudo-equilibrium within
Effluent NO3+NH4 the column.
The influent pH was between 7.8 and 8.2 during the
25 tests, and the DO was B8 mg/L. The effluent pH from
Concentration (mg-N/L)

the FE5 column had more variability (range: 7.8–9.8;

20 average=8.670.5; n ¼ 26) than from the FE7 column
(range: 8.0–8.4; average=8.270.1; n ¼ 27). The effluent
DO from both columns were very similar, and decreased
15
from 3 to 4 mg/L at the beginning of the test to less than
1 mg/L (0.670.2; n ¼ 40) after approximately 80 BV.
10 Grab samples immediately analyzed for ferrous iron
indicated that greater than 90% of the iron in the
packed column effluents was ferric iron. This was
5
surprising given the low DO and reducing conditions
in the column. Iron concentrations exiting the columns
0 increased, from B0.8 mg/L in the influent water to
0 100 200 300 400 500 600 700 roughly 5–6 mg/L during initial operation (Fig. 5). Iron
BV Treated concentrations gradually decreased from 5–6 to
Fig. 4. Nitrate-N reduction across a laboratory column with o1.5 mg/L over the first 500 BVs. The decline in iron
FE7 during extended operation, and ammonium-N production release as a function of BV roughly correlated with the
plus the sum of ammonium-N and nitrate-N in the column lack of a nitrogen mass balance closure (Figs. 3 and 4).
effluent (ambient groundwater with nitrate spiked to B25 mg Poor nitrate removal observed between 550 and 600 BVs
N/L). (Figs. 3 and 4) was accompanied by an increase in iron
release (up to 2 mg/L for FE5 and 8 mg/L for FE7). This
suggested that nitrate removal may be related to the
than observed in the short-term continuous flow lab presence and stability of iron coatings on the Fe0, and
tests which averaged approximately 70% recovery of that slight changes in initial water quality could disturb
nitrate as ammonium (Fig. 2). It was possible that such established iron coatings.
1824 P. Westerhoff, J. James / Water Research 37 (2003) 1818–1830

3.3. Field column operation closure on the nitrogen mass balance was especially
noticeable during Run 2, which had higher influent
3.3.1. Nitrate removal and by-products nitrogen levels (all nitrogen was nitrate).
The field Fe0 packed bed columns removed greater Effluent from the Fe0 packed bed had a higher pH
than 80% of the influent nitrate throughout the testing (7.5–9.5) than the influent pH (6.5–7.5). The field pH
(average=95% removal). Run 1 was conducted with was lower than laboratory experiments with the same
ambient nitrate levels in the groundwater (2–5 mg N/L; water due to degassing of carbon dioxide and equilibra-
Fig. 6). The variation in the groundwater was perhaps tion with the atmosphere during laboratory experiments.
expected, since the well serves as a hydrologic linkage A larger differential in pH (influent vs. effluent) was
between two different aquifer units. Localized pumping observed during the first half of both Runs 1 and 2,
affects level in the aquifer units, which have different compared against the change in pH during the later
nitrate levels. Hence, localized pumping affected influent stages of each test. During the later stages of the column
nitrate levels during the tests. Run 2, with a freshly operation less than a 0.5 pH unit increase across the
packed bed, was intended to be run at higher initial column was observed.
nitrate levels by spiking sodium nitrate in-line via a DO measurements indicated lower DO in the column
peristaltic pump. Varying flowrates through the Fe0 effluent (o2 mg/L) than the influent (4–5 mg/L). The
made pacing the nitrate feed pump difficult, and during effluent DO ranged from o0.2 to 4 mg/L, but averaged
Run 2 the influent nitrate levels varied between 5 and 1.871.1 mg/L. Sampling the effluent DO was difficult,
16 mg N/L. so the reported values represent qualitative trends only.
The first hundred BVs of the field column resulted in Iron release was monitored during Run 2. Effluent
approximately 70% conversion of nitrate removed into from the Fe0 packed bed had variable, but always
ammonium produced (Fig. 6). These observations were higher, soluble iron levels (average=2.4 mg/L) than the
consistent with the shorter-term laboratory column influent source water. Iron precipitates were visible (i.e.,
tests. Beyond the first few hundred BVs, lower nitrogen red color) in the effluent tubing and at the effluent waste
recoveries were observed for influent nitrate as nitrate disposal location. High variability in effluent iron
and ammonium (nitrite was always o0.1 mg/L). Lack of measurements may have been associated with sloughing
of this material during sampling. The effluent iron
5
concentrations were comparable to levels observed in
Run #1 the laboratory column tests.
4
Concentration (mg-N/L)

3.3.2. Transformations in Fe0 structure after continuous

3
operation
At the ends of both Runs 1 and 2, when permeability
decreased to levels o10% of initial flowrates, the
2
columns were returned to the laboratory and disas-
Influent NO3
sembled for visual observation and quantitative analysis.
1
Effluent NO3 The columns had been operated in an upflow mode.
Effluent NH4
Visible red solids were present in the lower 20% of the
0
0 500 1000 1500 2000 2500
column. Small (o1 mm) clumps of white material were
Cumulative BV Treated
present at the column inlet. The bottom 10 cm of the
100 cm long column (influent side) consisted of highly
20
Run #2
cemented iron and sand material. A hammer and chisel
Influent NO3
Effluent NO3 were later required to break the material apart. The
Effluent NH4 intermediate sections after Run 2 (above the iron oxides,
Concentration (mg-N/L)

15
but within the zone with 50% FE5 and 50% sand) were
black in color and appeared visually similar to the
10 original FE5 material. However, the irregular shape of
individual Fe0 particles had been lost. In contrast, iron
5
samples from the upper half of the column had
maintained the irregular shape of the iron but had
changed to a black–gray color. The material from the
0 intermediate and upper zones of the column was not
0 200 400 600 800 1000 1200
cemented together like the material at the bottom of the
Cumulative BV Treated
column (i.e., column influent). Visual observation of the
Fig. 6. Transformation of nitrogen-N species across the field column material from both Runs 1 and 2 were similar.
FE5 column for the period of operation for Runs 1 and 2. Approximately 20 samples were collected from various
 P. Westerhoff, J. James / Water Research 37 (2003) 1818–1830
(A) 1 mm
Fig. 7. SEM image of (A) virgin/unused FE5 and (B) iron material recovered from the top 10 cm of the field column
locations in the column from Run 1 and analyzed by
SEM and EDX analysis.
SEM images of virgin/unused FE5 material and of a
sample from field column (top 10 cm—exit end) are
shown in Fig. 7. The irregular shape of the iron shavings
was lost due to presumably to corrosion (i.e., iron
solubilization). EDX analysis of the virgin/unused FE5
material is presented in Fig. 8A, and indicates nearly
pure iron (Au from gold leaf coating).
EDX analysis and SEM images from the bottom
10 cm (influent side) of the field column with FE5
indicated the presence of iron cementous material and
calcium deposits. Fig. 8B and C indicate the presence of
calcium deposits on the iron, whereas calcium was
absent from the virgin FE5 (Fig. 8A). The SEM image
shown in Fig. 9A suggested that the iron has a coated
surface (EDX analysis for sample in Fig. 8B). Fig. 9B
shows a cement material that joined two iron particles
together. By EDX analysis the cement material was
comprised mostly of iron, with smaller calcium colloidal
deposits attached.
EDX analysis of the top 10 cm of the column
indicated the presence of iron platelets and other metal
deposits. Fig. 9C shows the presence of platelets, and
EDX analysis confirmed they were mostly iron (not
shown). EDX analysis revealed that deposits on the iron
platelets contained calcium and other metals (Fig. 8D).
4. Discussion
4.1. Nitrate removal
Zero-valent iron reduced nitrate concentrations in
packed columns, with the amount of removal dependent
upon the hydraulic contact time, initial nitrate concen-
tration, initial DO, and initial pH. Nitrate removal was
1825
(B) 1 mm
accompanied by ammonium production, proton con-
sumption, DO consumption, and soluble iron produc-
tion. Nitrite was not observed. Rapid nitrite reduction
on iron surfaces (e.g., green iron rust) may explain an
absence of nitrite accumulation [21].
The iron chips Fe0 material (FE7) removed less nitrate
than the iron filing Fe0 material (FE5), and was likely
due to the 75 times higher surface area of FE5 (1.2 m2/g)
than FE7 (0.016 m2/g). For both Fe0 material increasing
the hydraulic residence time in the packed columns
increased the nitrate removal rate, which varied from
o1 to >18 mg NO3-N/L-reactor/h. Nitrate removal
rates increased with increasing initial nitrate concentra-
tions, but by different extents for the two Fe0 materials.
For reference, the order of magnitude of the nitrate
removal rates were comparable to those observed during
autotrophic denitrification or anaerobic wastewater
treatment systems (1–10 mg NO3-N/L-reactor/h) [22,
23].
Increasing initial DO concentrations lead to increased
percentage nitrate removals across the Fe0 packed
columns. This observation was contrary to batch
experiments where DO inhibits nitrate reduction. A
possible explanation could involve the formation of
unstable surface iron precipitates (e.g., green iron rusts),
produced in the presence of oxygen, which may play a
role in nitrate reduction [12, 24]. Batch experiments have
higher solid to liquid ratios than packed columns. As a
consequence, any modifications in surface chemistry due
to oxygen reactions with Fe0 surfaces may be minor in
batch experiments, compared to larger impacts in
packed columns. Packed columns also have a contin-
uous loading of DO, allowing modifications of the Fe0
surface or precipitates to form.
Decreasing the initial pH increased nitrate removal,
and pH increased across the Fe0 packed bed. The
linkage between nitrate removal and pH can be
1826 P. Westerhoff, J. James / Water Research 37 (2003) 1818–1830

4000 Fe
Au
3000

Counts
2000

Si
1000 Fe
Au
Fe Au
0
0 5 10 15
(A) Energy (keV)

1200
Fe
1000
800
Counts

600 Au

400 Si
Ca
200 Fe Au
Au
Ca
0
0 5 10 15 20 25
(B) Energy (keV)

200
Fe
150
Counts

Ca
100

50
SiAu Fe
Ca
0
5 10 15 20 25
(C) Energy (keV)

6000 Fe

5000

4000
Counts

3000

2000 Au
Si Mn Fe
1000 Mn Ca Cr Au
Cr Au
0
0 5 10 15
(D) Energy (keV)

Fig. 8. X-ray image from selected SEM images for FE5 field column: (A) virgin/unused material, (B) from bottom 10 cm of column
(SEM shown in Fig. 9A and B), (C) duplicate sample from bottom 10 cm of column, and (D) from the top 10 cm of column (SEM
shown in Fig. 9C).
 P. Westerhoff, J. James / Water Research 37 (2003) 1818–1830 1827

(A) 100 um (B) 20 um

(C) 50 um

Fig. 9. SEM images at (A) 160  magnification for a sample from the bottom 10 cm of the column (X-ray count in Fig. 8B), (B)
1000  magnification of cemented material from the bottom 10 cm of the column, and (C) 350  magnification of platelets from the
top 10 cm of the column.

described by the reaction stoichiometry where 10 mol of
protons were consumed for every mole of nitrate
converted to ammonium ion:
3 þ 10H þ 4Fe -NH4 þ 3H2 O þ 4Fe :
NO þ 0 þ 2þ
ð3Þ
Experiments with larger mass nitrate removals (i.e.,
higher initial nitrate, longer EBCTs) resulted in a larger
change in pH. This implies that corrosion was necessary
for nitrate reduction. Soluble iron concentrations exiting
the laboratory columns averaged 2.3 mg/L for FE5 and
8.5 mg/L for FE7. The fate of corroded iron (soluble or
precipitates) may be critical to the feasibility and
sustainability of Fe0 systems. The secondary MCL for
iron is 0.3 mg/L. Mobilized iron exiting the packed bed
was mostly present as ferric iron. Ferrous iron may have
reacted with nitrate, or oxygen, to form ferric iron or
precipitated as a new mineral [17,25]. The role of iron
and other precipitates is discussed below. In general,
increasing pH due to proton consumption (Eq. (3)) will
decrease iron solubility and allow precipitation of iron
solids.
In comparison of lab- and field-scale Fe0 columns,
closures of nitrogen mass balances were observed during
initial phases (200–300 BVs) of the extended-run lab-
scale (Figs. 3 and 4) and field-scale (Fig. 6) columns.
After the first 200–300 BVs, the nitrogen mass balances
were not closed by accounting for nitrate, nitrite, and
ammonia. A larger discrepancy between influent and
effluent nitrogen concentrations was observed for
the field-scale column compared to the lab-scale
column. Near complete removal of nitrate was observed
in the field columns, which over time had lengthy
1828 P. Westerhoff, J. James / Water Research 37 (2003) 1818–1830
hydraulic residence times due to flow restrictions
imparted by precipitated materials. In contrast, batch
experiments with Fe0 and nitrate always lead to
near complete conversions of nitrate to ammonia and
closure of a nitrogen mass balance [8–13,18]. This
suggests that batch experiments may not be adequate
in predicting performance of continuous flow packed
columns, or permeable subsurface barriers. The
results do suggest that most features of field
operation can be simulated by longer-duration labora-
tory column.
4.2. Formation of precipitates in columns
A primary difference between the field- and lab-scale
column tests was the more significant reduction in
hydraulic permeability through the field column, experi-
enced after 1000 BV. This difference may be attributed
to water chemistry differences in water transported upon
delivery to the laboratory, as opposed to continuous
supply of the field column directly from the ground-
water. Particle deposition would primarily impact flow
at the entrance through the column. However, solidified/
cemented material in the column was unlikely due to
influent sediment. The field pH was slightly lower than
in the lab experiments due to carbon dioxide degassing,
resulting in more corrosive field conditions. Further-
more, the field column had higher inorganic carbon
concentrations which allow the formation of carbonate
precipitates.
SEM results suggested a morphological change
occurred of the Fe0 over time, resulting in a smoother/
dissolved surface, but also resulted in the precipitation
of iron oxide/carbonate and calcium-rich deposits that
cemented the Fe0 particles together. Cementation within
the packed column decreased the permeability, ulti-
mately requiring high influent water pressures to achieve
any flow through the column.
The visibly red oxide layer in the columns contained
both iron and calcium precipitates. Others have reported
aragonite (CaCO3(s)) formation in situ for a ground-
water treatment Fe0 system with high calcium hardness
groundwater [17]. Waters containing elevated levels of
bicarbonate and sulfate are particularly corrosive to Fe0
columns, resulting in the formation of ferrous carbonate
compounds [26]. MacKenzie et al. [27] suggested that
Fe(OH)2, FeCO3, and CaCO3 in higher carbonate
waters would begin to dominate porosity losses. For-
mation of several metal precipitates has been reported
[28], and given the relatively high sulfate and carbonate
concentrations in our water (Table 1) it was likely that
iron hydroxide, aragonite, and green iron rust could
form. Based upon SEM and EDX analysis the
precipitates form on the Fe0 surface and in pores
between Fe0 particles, increasing the surface area but
decreasing the porosity.
4.3. Impact of precipitates on nitrate reduction
During the continuous operation of Fe0 packed
columns for hundreds to thousands of BVs, changes in
the performance were observed. Initially, the field
columns behaved similarly to the laboratory columns
and were characterized by the following trends: (1)
reducing nitrate to mostly ammonium; (2) increasing pH
across the column; (3) decreasing DO across the column;
(4) solublization of iron across the column; and (5)
maintenance of flow characteristics (permeability). Dur-
ing the second half of both Runs 1 and 2 different
patterns were observed, including the following: (1)
nitrate reduction occurred but a low recovery as
ammonium was observed; (2) less of an increase in pH
occurred across the column; (3) permeability decreased
across the column; and (4) iron oxide production was
qualitatively observed (visually and by SEM). Elucida-
tion of nitrate removal mechanisms in the presence of
Fe0 was not a primary focus of this study, but the results
can be interpreted to develop multiple hypotheses that
could guide future research. Plausible abiotic and biotic
mechanisms for observed patterns in nitrate removal
and nitrogen mass balance closure are discussed below.
Four abiotic factors may have influenced the observed
nitrogen dynamics. First, the lack of a nitrogen mass
balance closure was influenced by ammonia (NH3)
volatilization at elevated pH (>9). While plausible,
careful sampling techniques should have avoided this
mechanism, but it could occur in some field settings.
Second, the lack of a nitrogen mass balance closure was
influenced by ammonium ion (NH+ 4 ) or nitrate sorption
onto precipitated metal oxides. For example, iron oxides
have pH zero-point-of-charges between 5.0 and 8.5, thus
in the pH range observed in the lab and field columns
(7.5–9.0) many surface oxides would have net negative
surface charges and allowed ammonium ion adsorption
to occur. Ammonium ions can be strongly adsorbed/
exchanged in soils and on iron oxides [29–32]. Further-
more, it was observed in a limited number of laboratory
experiments that pulses of nitrate were observed exiting
the column between shorter-term experiments, suggest-
ing that nitrate adsorbed onto the mineral precipitates.
Nitrate sorption could occur via sorption onto positively
charged surfaces through electrostatic attraction or
ligand exchange (i.e., surface hydroxide; green iron rust)
[24,33]. The apparent lack of closure in the nitrogen
mass balance over time could have occurred due to
sorption of nitrate (or ammonium ion) onto mineral
deposits that formed over time in the column. Third, un-
monitored nitrogen gas species (e.g., NO2, N2O, N2H4)
although not significant in previous batch experiments
[13,18], may have been produced in the columns.
Unstable intermediates (e.g., NH4NO2, NH4NO3) may
be produced although evidence of them in Fe0 systems
has not been verified [34,35]. Fourth, iron corrosion by-
 P. Westerhoff, J. James / Water Research 37 (2003) 1818–1830
products form, increase the solids surface area in the
packed column over time, and may alter the mechanisms
of electron transfer, nitrate reduction, and nitrogen by-
products. Nitrate can be stoichiometrically reduced to
ammonium in the presence of green iron rust com-
pounds (FeII III
4 Fe2 (OH)12SO4  yH2O), and lead to the
production of magnetite as an oxidation by-product
[12,24]. Water used in the current work (Table 1)
contained moderately high levels of sulfate, and green
iron rust may have formed within the packed bed.
Cheng et al. [13] hypothesized that iron corrosion by-
products (e.g., Fe(+II)aq, Fe(OH)2(s)), produced by
reactions of Fe0 with oxygen, were thermodynamically
capable of reducing nitrate to ammonia. Scherer et al.
[32] described the formation of oxide layers on Fe0, and
how such formation could affect electron transfer
between the metal and dissolved ions/ molecules.
Several biological-mediated hypotheses could also be
used to support the observations, although no measure-
ments of biological activity were conducted as part of
the present study, and a review is available elsewhere
[36]. First, heterotrophic denitrification was not a
primary nitrate removal mechanism. The groundwater
contained low organic carbon content (o0.2 mg/L).
Even if the organic carbon was present in a readily
biodegradable form (e.g., acetate), it was present in
insufficient concentrations to remove the applied nitrate
(5–50 mg N/L). Second, autotrophic bacteria may have
became established over time in the columns and
participated in nitrate reduction. Straub and others
reported a group of bacteria were found that coupled the
oxidation of Fe2+ to a complete denitrification process
[37,38]. Microbially (autrophic) catalyzed nitrate-depen-
dent ferrous iron oxidation for several iron solids (e.g.,
iron oxides and carbonates) has been observed [39],
providing a basis to investigate microbial nitrate
reduction processes. Soares [40] reviewed heterotrophic
and autotrophic denitrification of groundwaters, includ-
ing remediation processes involving hydrogen gas and
various reduced sulfur-compounds. The latter would not
likely be significant in the packed columns used in the
current study. Chemoautotrophs using H2 can reduce
nitrate in the presence of Fe0 [41]. Although such
reactions are typically mediated by low pH, the presence
of aluminosilicate minerals can increase the rate of oxo-
anion reduction by inducing localized surface produc-
tion of protons that may promote H2 formation [42].
Alternatively, given that ammonium was a nitrate
reduction by-product, the presence of both nitrate and
ammonium may have facilitated establishment of
anaerobic ammonium-oxidizing (AMMONOX) bacteria
that have the following stoichiometry for nitrogen
4 +3NO3 -4N2+9H2O+2H .

production [43]: 5NH+ +
Ammonium ion sorption onto metal oxides could
provide localized high ammonium concentrations avail-
able for the AMMONOX bacteria.
1829
Since nitrate reduction requires corrosion of iron and
corresponding increase in pH, it was uncertain whether
nitrate removal could be attributed solely to abiotic
corrosion mechanisms. Future work should address the
role of ammonium ion sorption and role of autotrophic
bacteria on the enhanced longer-term nitrogen removal
performance of Fe0 systems. However, the limiting
factor for Fe0 system sustainability will likely be related
to the reduction in permeability and formation of
passivating layers that may limit mass transport of
nitrate to the reactive surface. It may be possible to re-
establish initial conditions by chemical treatments (e.g.,
acid injection).
5. Conclusions
Incorporation of zero-valent metals (e.g., Fe0) and/or
metallic catalysts into water treatment processes will
also have to address the production of by-products (e.g.,
ammonium) from the oxo-anions, from the metal (e.g.,
dissolved iron), and from indirect reactions (e.g., rise in
pH due to proton consumption). Indirectly, the presence
of oxo-anions (e.g., nitrate, chlorate) in drinking water
may also affect, and be affected, by contact with metals
during water treatment, storage, and distribution. Oxo-
anions may oxidize metals (e.g., iron pipe in distribution
systems, copper pipe in households), be chemically
reduced, release metal by-products into the water (e.g.,
dissolved iron or copper), or affect solution pH. The
work presented herein provides a framework for future
studies to address these issues.
Acknowledgements
This work was supported by American Water Works
Association Research Foundation (Denver, CO) under
the supervisor of Traci Case (Project Officer). Sugges-
tions from the Project Advisory Committee (Pedro
Alvarez, Tian Zhang, and Edwin Guasp) were
appreciated.
References
[1] Cantrell KJ, Kaplan DI, Wietsma TW. J Hazard Mater
1995;42:201–12.
[2] Cantrell KJ, Kaplan DI. J Environ Eng—ASCE
1997;123:499–505.
[3] Blowes DW, Ptacek CJ, Jambor JL. Environ Sci Technol
1997;31:3348–57.
[4] Farrell J, Melitas N, Kason M, Li T. Environ Sci Technol
2000;34:2549–56.
[5] Chew CF, Zhang TC. Water Sci Technol 1998;38:135–42.
[6] Powell RM, Puls RW, Hightower SK, Sabatini DA.
Environ Sci Technol 1995;29:1913–22.
1830 P. Westerhoff, J. James / Water Research 37 (2003) 1818–1830
[7] Zawaideh LL, Zhang TC. Water Sci Technol 1998;38:
107–15.
[8] Murphy AP. Nature 1991;350:223–5.
[9] Siantar DP, Schreier CG, Chou CS, Reinhard M. Water
Res 1996;30:2315–22.
[10] Huang CP, Wang HW, Chiu PC. Water Res 1998;32:
2257–64.
[11] Daum J, Vorlop KD. Chem Eng Technol 1999;22:199–202.
[12] Hansen HCB, Koch CB, NanckeKrogh H, Borggaard OK,
Sorensen J. Environ Sci Technol 1996;30:2053–6.
[13] Cheng IF, Muftikian R, Fernando Q, Korte N. Chemo-
sphere 1997;35:2689–95.
[14] Fan AM, Stienberg VE. Regul Toxicol Pharmacol
1996;23:35–43.
[15] Andrews D, Frosard W, Plummer AH, Mokry LE. Lake
Reservoir Manage 1994;9:63.
[16] Kapoor A, Viraraghavan T. J Environ Eng—ASCE
1997;123:371–80.
[17] Phillips DH, et al. Environ Sci Technol 2000;34:4169–76.
[18] Westerhoff P, Environ Eng—ASCE, accepted for publica-
tion.
[19] Levenspiel O. Chemical reaction engineering. New York,
NY: Wiley, 1999.
[20] AWWA APHA, & WEF. Standard Methods for the
Examination of Water and Wastewater, 19th ed. American
Public Health Association, 1995.
[21] Hansen HCB, Borggaard OK, Sorensen J. Geochim
Cosmochim Acta 1994;58:2599–608.
[22] Metcalf & Eddy Inc. In: Tchobanoglous G, editor.
Wastewater engineering: treatment disposal reuse, 2nd
ed., revised. New York, NY: McGraw-Hill, 1979.
[23] Szekeres S, Kiss I, Bejerano TT, Soares MIM. Water Res.
2001;35:715–9.
[24] Hansen HCB, Koch CB. Clay Miner 1998;33:87–101.
[25] Ottley CJ, Davison W, Edmunds WM. Geochim Cosmo-
chim Acta 1997;61:1819–28.
[26] Gu B, Phelps TJ, Liang L, Dickey MJ, Roh Y, et al.
Environ Sci Technol 1999;33:2170–7.
[27] Mackenzie PD, Horney DP, Sivavec TM. J Hazard Mater
1999;68:1–17.
[28] Yabusaki S, Cantrell K, Sass B, Steefel C. Environ Sci
Technol 2001;35:1493–503.
[29] NunezDelgado A, LopezPeriago E, DiazFierrosViqueira
F. Water Res 1997;31:2892–8.
[30] Thornton SF, Tellam JH, Lerner DN. J Contam Hydrol
2000;43:327–54.
[31] Ceazan ML, Thurman EM, Smith RL. Environ Sci
Technol 1989;23:1402–8.
[32] Scherer MM, Balko BA, Tratnyek PG. In: Grundl, editor.
ACS Symposium Series 715: Mineral-Water Interface
Reactions Kinetics and Mechanisms. Washington, DC:
American Chemical Society, 1999.
[33] Randall SR, Sherman DM, Ragnarsdottir KV. Geochim
Cosmochim Acta 2001;65:1015–23.
[34] Takenaka N, Suzue T, Ohira K, Morikawa T,
Bandow H, Maeda Y. Environ Sci Technol 1999;
33:1444–7.
[35] Trogler WC. Coordin Chem Rev 1999;187:303–27.
[36] Scherer MM, Richter S, Valentine RL, Alvarez PJJ. Crit
Rev Microbiol 2000;26:221–64.
[37] Straub KL, Benz M, Schink B, Widdel F. Appl Environ
Microbiol 1996;62:1458–60.
[38] Straub KL, Buchholz-Cleven BEE. Appl Environ Micro-
biol 1998;64:4846–56.
[39] Weber KA, Picardal FW, Roden EE. Environ Sci Technol
2001;35:1644–50.
[40] Soares MIM. Water Air Soil Pollut 2000;123:
183–93.
[41] Till BA, Weathers LJ, Alvarez PJJ. Environ Sci Technol
1998;32:634–9.
[42] Powell RM, Puls RW. Environ SciTechnol 1997;31:
2244–51.
[43] Mulder A, Vandegraaf AA, Robertson LA, Kuenen JG.
Fems Microbiol Ecol 1995;16:177–83.
[44] S. R. Project, Annual Water Quality Report. Salt River
Project—Operational Support Division, 1993.