Identification and Quantification of Min

Environ. Sci. Technol. 2004, 38, 5757-5765

Identification and Quantification of
Mineral Precipitation in Fe 0 Filings
from a Column Study
WIWAT KAMOLPORNWIJIT,†
L I Y U A N L I A N G , * ,†,‡
GERILYNN R. MOLINE,†
TODD HART,§ AND OLIVIA R. WEST†
Environmental Sciences Division, Oak Ridge National
Laboratory, P.O. Box 2008, Oak Ridge, Tennessee 37831-6036,
Cardiff School of Engineering, Cardiff University, Queen’s
Buildings, P.O. Box 925, Cardiff CF24 0YF, Wales, U.K., and
Battelle Pacific Northwest National Laboratory, P.O. Box 999,
Richland, Washington 99352

Thermogravimetric analysis (TGA) combined w ith X-ray
diffraction (XRD) w as used to identify mineral phases and
determine corrosion rates of granular iron samples from
a 2-yr field column study. Similar to other studies, goethite,
magnetite, aragonite, and calcite w ere found to be the

major precipitated minerals, w ith Fe 2(OH)2CO3 and green
rust as minor phases. Based on TGA-mass spectrometry
(M S) analysis, Fe 0 corrodes at rates of 0.5-6.1 mmol kg -1
d -1 in the high NO3- (up to 13.5 mM ) groundw ater; this
rate is significantly higher than previously reported. Porosity
reduction w as 40.6%-45.1% for the inlet sand/Fe 0 interface
and 7.4%-25.6% for effluent samples of tw o test columns.
Normalized for treatment volumes, porosity loss values are
consistent w ith studies that use high levels of SO42- but
are higher than those using low levels of corrosive species.
Aqueous mass balance calculations yield corrosion
rates similar to the TGA-M S method, providing an alternative
to coring and mineralogical analysis. A severely corroded
iron sample from the column simulating a 17-yr treatment
throughput show ed >75% porosity loss. Extensive porosity
loss due to high levels of corrosive species in groundw ater
w ill have significant impact on long-term performance
of permeable reactive barriers.
Introduction
Zerovalent iron (Fe0) has been widely used as a reactive

medium in permeable reactive barriers (PRB) for treating
organic (11) and inorganic contaminants (12). Knowledge of
rates and products of Fe0 corrosion and other biogeochemical
reactions associated with long-term water flow within the
PRB is of critical importance for efficient application of this
technology (3-5, 13). Precipitation of secondary minerals,
defined as phases not involved in containment of the
contaminant (1), can lead to loss of iron reactivity (13),
reduced hydraulic conductivity (1, 14), and the development
of preferential flow (15, 16). The effects of these on the longterm performance of a PRB will depend on the type of
minerals precipitated as well as the mass of precipitation
over time. A wide range of mineral precipitates has been
* Corresponding author phone: +44(0)29 2087 6175; e-mail:
Liang@cardiff.ac.uk.
† Oak Ridge National Laboratory.
‡ Cardiff University.
§ Battelle Pacific Northwest National Laboratory.
10.1021/es035085t CCC: $27.50
Published on Web 09/29/2004

 2004 Am erican Chem ical Society

identified in laboratory and field studies, including amorphous iron oxyhydroxide (ferrihydrite), aragonite, calcite,
crystalline iron (oxyhydr)oxide (akaganeite, goethite, magnetite, hematite, lepidocrocite, etc.), green rust, siderite, and
mackinawite (1-5, 8, 10). However, direct methods for
quantifying mineral precipitates and the long-term impact
of these corrosion products on Fe0-PRB longevity are not
well established.
Precipitates can be quantified using mass balance of pore
water chemistry up- and down-gradient of a PRB, assuming
that such precipitates (such as CaCO3) are the sink of aqueous
species (9, 14, 17). Using mass balance approach to quantify
iron-bearing precipitates, however, requires knowledge of
iron corrosion rates because dissolved Fe from corrosion is
concurrently removed by precipitation (9, 18, 19) within the
PRB. Published values of Fe0 corrosion rate are sparse (6, 20)
and are not applicable to all types of iron and groundwater.
Direct XRD quantification of iron-containing precipitates in
core samples suffers from overwhelming interference of Fe0
that cannot be completely separated from corrosion products.

Sequential chemical extraction, a common method for soil
mineral analysis, is not suitable for Fe0 samples because Fe0
dissolves in the extraction solution. An effective technique,
therefore, must either permit the analysis of precipitates
without separation from the Fe0 or be unaffected by the
presence of Fe0.
Thermogravimetric analysis (TGA) has been widely used
in the study of phase transformations at different temperatures (21). In this technique, mineral phases are quantified
by integrating weight changes during transformation. The
use of mass spectrometry (MS) with TGA for off-gas analysis
can confirm the presumed phase. Because Fe0 is thermodynamically stable under inert or reducing conditions, it does
not influence TGA results. With knowledge of phase stability
and transformation at given temperatures, identification and
quantification of some important mineral phases can be
made without having to extract or separate precipitates from
the Fe0 media.
In this study, quantification of precipitates and derivation
of Fe0 corrosion rates were pursued using core samples from
field columns subjected to ∼2-yr treatment of groundwater
containing up to 13.5 mM NO3- (16). TGA-MS combined

with X-ray diffraction (XRD) and scanning electron microscopy (SEM) were used to directly quantify precipitate
compositions in iron core samples. Degrees of cementation
and pore filling within Fe0 were estimated for samples treated
under both accelerated- and normal-flow conditions. This
method is compared with a mass balance approach which
uses pore water composition to estimate Fe0 corrosion and
precipitation. The impact of mineral precipitation on the
performance of the PRB in terms of porosity reduction is
assessed.

Experimental Methods
Materials. Iron core samples were collected from two large
field columns (92-cm long, 15-cm diameter, see Plate 1,
Supporting Information) that were operated for ∼2 yrs at the
Y-12 National Security Complex in Oak Ridge, Tennessee
(16). The columns were oriented horizontally ∼70-ft upgradient from an existing PRB (Figure 1 in ref 22), using the
same iron filings as the PRB. Two flow rates were used,
yielding initial pore velocities of 9.4 m d-1 in Column I and
0.33 m d-1 in Column II. The slower flow was intended to
simulate typical field conditions, while the faster flow was

designed to simulate treatment of high throughput for a
period of ∼17 years. Insufficient residence time may not
VOL. 38, NO. 21, 2004 / ENVIRONM ENTAL SCIENCE & TECHNOLOGY

9

5757

FIGURE 1. Chemical changes along Column I (a and b) and Column II (c and d), show ing pH (]), calcium (0), nitrate (4), and ferrous
iron (O) in pore w ater along centerline. Steady state change w as observed during early operation (a and c), exhibiting homogeneous flow .
Heterogeneous flow appeared during later stage operation (b and d), w here higher concentrations at dow n gradient location are due to
mixing w ith preferentially transported influent (16). Dash lines show the interfaces of iron and sand, w ith Fe 0 sandw iched betw een tw o
layers of sands.

adequately simulate high volume treatment if biogeochemical
reactions do not reach completion. Nonetheless, the columns
do appear to have captured major corrosion and precipitation
reactions (16, 17) as indicated by high effluent pH and
progressive removal of Ca and NO3- in pore water (Figure
1a,c) during the first two-month operation. Site groundwater

was directly pumped into the columns; the composition
varied due to seasonal changes, exhibiting 6.6-7.4 pH, 3.76.4 mM alkalinity, 5.4-9.0 mM Ca, 0.8-13.5 mM NO3-, 0.60.9 mM SO42-, 1.0-2.5 mM Cl-, and ∼9 µM dissolved O2.
Figure 1 shows several snapshots of chemical profiles along
the columns: homogeneous flow condition during the early
stage (Figure 1a,c) and heterogeneous at later stage (Figure
1b,d). Description of column set up, operation, and the
geochemical and hydrologic monitoring results are presented
in detail elsewhere (16, 17).
Column I was disassembled after 680-d operation, having
treated 11.5 m3 groundwater (1336 pore volume). Pore
volumes were calculated based on an initial porosity
measured at 60% (16). Dissection required use of a hole-saw
for the collection of core plugs (2.9-cm diameter by 4.1-cm
long) under atmospheric conditions. Exposure time was
minimized by immediate transfer of samples to plastic bags,
which were stored in a closed PVC tube flushed with Ar gas
at 5-psi pressure. The column was dissected in ∼5-cm
sections, typically with 5 cores from each section (Figure 2,
also see Plate 2, Supporting Information).
Column II was disassembled after 695 d, having treated

1.1 m3 groundwater (∼150 pore volume). Sample collection
was done easily with a spatula. Samples were collected and
stored in Ar as done for Column I samples.
Sample Preparation and Analysis. Most of the iron core
samples from Column I maintained the cylindrical shape of
5758

9

ENVIRONM ENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 21, 2004

the hole-saw. The surface of the core plug may be affected
by drilling-induced heating and atmosphere exposure and
was therefore removed prior to sampling for mineralogical
analysis. Subsamples were used for XRD analysis and TGAMS and SEM characterization. The XRD samples were
sonicated for 1/2 hour in acetone and then wet ground
under Ar flow. Upon drying, samples were sieved through a
#120 (125 µm) sieve and the finer fractions collected for
XRD (Scintag XDS2000) analysis. Samples for TGA-MS and
SEM analysis were dried anoxically before further preparation.

TGA-MS analysis was performed on selected samples,
including the following: (1) 0-4 cm from the inlet sand-Fe0
interface (CS-I-8, CS-II-1), (2) mid-span (CS-I-81, 37-41 cm
into Column I; CS-II-3, 37 cm into Column II), and (3) 0-4
cm from the effluent sand-Fe0 interface (CS-I-49, CS-II-15).
The term “CS” denotes “column sample”, roman numerals
identify the columns, and integers identify samples (Figure
2). Stock iron filings were also analyzed. Samples (∼100 mg)
were placed in pure alumina crucibles and heated to 900 °C
at a rate of 10 °C min-1 under He gas flow at a rate of 100
mL min-1. Analyses were performed using a Netzsch STA
409 TGA/DSC and a Pfeiffer QMS300 MS. After reaching 900
°C, 2% H2 gas in He was passed through the furnace causing
transformation of iron oxides to Fe0.
A dry sample from Column I (CS-I-14) was prepared for
SEM analysis. It was impregnated in Buehler Epoxide resin
and hardener and subjected to vortexing for 30 s. The
solidified sample was cut, polished, and dried anoxically
before surface carbon coating (PELCO, CC-7A). The thin
section was observed using Scanning Electron Microscope

FEI XL30FEG, and elemental analysis was done using an
energy dispersive detector (OXFORD ultrathin window).

FIGURE 2. Phase identification in numbered samples along and across Columns I and II. The mineral phases w ere listed in the decreasing
order of XRD intensity. Calcite and goethite generally predominated over aragonite and magnetite near the influent and in regions adjacent
to the column gap. The shaded dark gray sections show iron media, w here cementation decreased from the influent to the effluent of
the column, betw een tw o sand layers. Phase abbreviations are used as follow s: G-goethite, C-calcite, A-aragonite, M -magnetite, FeFe 2(OH)2CO3, Gr-green rust, Q-quartz.

Results and Discussion
Visual Observation of Iron Corrosion. Column I. Iron
corrosion and cementation were extensive at the inlet sandFe0 interface and reduced progressively in the column. Near
the inlet, iron lost visible grains and showed heavy cementation (see Plate 2a, Supporting Information). Rusty stain was
visible throughout the iron cross-section and on sand grains
near the inlet. Cementation extended ∼60 cm into the
column, accounting for ∼2/3 of the iron mass. Distinct iron
grains were present ∼14 cm into the column, on otherwise
solidly cemented samples. In the same section, iron was more
easily broken up in the bottom part of the horizontal-oriented
column, implying less extensive cementation. Loose filings
mixed with cemented samples were observed at 22 cm, near

the bottom of the column, while the upper part of this section
was completely cemented. From 22 to 60 cm, however, the
bottom regions show heavy cementation. Rust stain was also
seen at these cross-sections with less intensity than at the
inlet interface (see Plate 2b, Supporting Information). Starting
at ∼60 cm, and extending to the effluent interface, loose but
densely compacted dark gray to black filings predominated.

Column II. With one tenth of the volume treated in
Column I, cementation and corrosion were substantially less
in this column than in Column I. Cemented samples were
obtained at the inlet sand-Fe0 interface, as in Column I.
However, cementation extended only ∼2.5 cm into the Fe0
accounting for ∼1/12 of iron mass. In the remainder of the
column, filings were loose with no observable cementation.
Rusty stain was observed at the influent and effluent sandFe0 interfaces.
For both columns, a 1-2 mm gap was observed between
the media and the column casing across the entire length of
the column, extending 6-7 cm in width along a ∼120 degree
arc (see Plate 2, Supporting Information). The surfaces of
the media facing this gap were completely coated with rusty
deposits, the products of Fe0 oxidation by O2. This implies
that the O2-rich influent was preferentially transported
through the gap. It is unknown when the gap formed. It may
have been initiated by settling and compaction of the iron
grains during initial setup and exacerbated by Fe0 corrosion
and grain size reduction. The gap itself, however, did not
necessarily cause flow channeling. Changes in pore water
VOL. 38, NO. 21, 2004 / ENVIRONM ENTAL SCIENCE & TECHNOLOGY

9

5759

TABLE 1. Precipitate Mass and Porosity Loss in Column I Based on the Mass Balance of Pore Water Aqueous Species, Modified
from Ref 16
porosity lossb
pore
volume

corrosiona rate
(mmol kg-1 d-1)

run time
day

simulated
time

30

2.1

163

17

72

5.1

398

10.1

215

7.2

569

3.1

399

12.4

974

4.5

666

16.6

1310

2.5

phases

precipitate mass
(mmol L-1)

entire
columnc

first
50 cmd

CaCO3
Fe3O4/R-FeOOH
total
CaCO3
Fe3O4/R-FeOOH
total
CaCO3
Fe3O4/R-FeOOH
total
CaCO3
Fe3O4/R-FeOOH
total
CaCO3
Fe3O4/R-FeOOH
total

3.6
5.4/16.3
9.0/19.9
1.8
1.9/5.6
12.7/27.3
0.9
1.4/4.1
15/32.3
1.7
1.5/4.6
18.2/38.6
1.0
1.0/2.9
20.2/42.5

2
2.1/5.1
4.1/7.1
2.4
1.7/4.2
8.3/13.8
1.8
1.8/4.5
12.0/20.1
5.7
3.5/8.6
12.9/20.5
4.3
3.0/7.3
20.1/32.1

3.3
3.5/8.5
6.8/11.8
4
2.9/7.0
13.7/22.9
3
3.0/7.4
19.8/33.3
9.4
5.8/14.2
35.1/56.9
7.1
5.0/12.1
47.1/76.1

a Assum ed abiotic nitrate reduction prior to day 215, and biotic reduction afterw ard: abiotic reaction: 4Fe0 + NO - + 7H O f 4Fe2+ + NH +
2
4
3
+ 10OH-; biotic reaction: 5Fe0 + 2NO3- + 6H2O f 5Fe2+ + N 2 + 12OH-. b The porosity loss w as calculated using 60% initial porosity. The porosity
0
loss due to Fe3O4 or R-FeOOH precipitation is separated by ‘/’ and has been corrected for any porosity gain from the loss of Fe m ass due to
corrosion. c Distributing m ineral phases uniform ly throughout the Fe0 m edium . d M ineral phases allocated to the first 50 cm of the iron m edium ,
w here heavy cem entation w as observed.

chemistry over time (Figure 1a,b) suggest gradual development of preferential flow paths (16, 17). Additionally,
hydraulic tracer tests in Column I showed homogeneous
flow after 1.3-yr of simulated normal treatment but severe
flow channeling after 3.9-yr simulation (16). Mass balance
estimates indicate that 9-27.3 mM of precipitates formed in
∼3-yr of simulated treatment time (run time of 30 and 72 d
in Table 1). We suggest that this precipitation at the inlet
region caused hydraulic conductivity contrast between the
gap and the bulk media and subsequently the diversion of
influent through the gap.
Following influent diversion, groundwater redistributed
beyond the heavily cemented region, reacting with Fe0 and
precipitating minerals nonuniformly into the column. The
reduced groundwater flow through the lower part of the
column may account for less cementation than in the top
region. Thus, heterogeneous flow appears to be the main
reason for the differing degrees of cementation across and
along the columns (16).
Mineral Phase Identification Using XRD. Quantitative
use of XRD intensity normally requires proper standardization. However, change in intensity ratio of certain phase pairs
implies change in mass ratio, provided samples have similar
matrix and phase assemblage. In this study, changes in
intensity ratios of two phase-pairs (calcite/aragonite and
magnetite/goethite) were observed to elucidate the influence
of pore waters on the phase formation. No attempt was made
to quantify phases using XRD.
Figure 2 summarizes the mineral phases identified in
cores, showing characteristic XRD signal intensities in
descending order. All samples showed a similar suite of
minerals but with different intensity ratios. The major phases
are goethite (R-FeOOH), magnetite (Fe3O4), calcite, and
aragonite, with lesser amounts of Fe2(OH)2CO3 and green
rust. These phases were identified under field and laboratory
conditions (1-5), with some exceptions. Siderite and sulfide
phases were not identified by XRD in this study but have
been reported in the nearby PRB (5, 7) and others (3). The
amorphous phases (iron oxide in particular), indicated by
broad peaks or raised background in XRD diffractograms,
were observed in field core samples (7) but not in this column
study. Furukawa (4) reported that XRD failed to detect
ferrihydrite, an amorphous phase, despite its ubiquitous
5760

9

ENVIRONM ENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 21, 2004

existence shown by selected area electron diffraction of field
core samples. The presence of an amorphous phase in the
columns may therefore be possible but to a much lesser
extent than the crystalline phases.
Distribution of predominant mineral phases in the
columns is consistent with groundwater characteristics and
known thermodynamic stability conditions for these phases
(19, 21, 23). In Column I calcite and goethite generally
predominated over aragonite and magnetite under conditions
of lower pH and higher dissolved O2, NO3-, Ca, and alkalinity
(i.e., near the influent region and adjacent to the column
gap, Figure 2). Elsewhere in the column this predominance
was reversed. In Column II goethite was detected only in
samples at the sand-Fe0 interface and regions adjacent to
the gap. The distribution of iron (oxyhydr)oxide is consistent
with previous findings that goethite precipitates in waters
containing dissolved O2 and lower pH, while magnetite is
the predominant phase at alkaline pH and higher Fe2+ (23).
Nonetheless, both goethite and magnetite may coexist when
there is an oxidation potential gradient across the corrosion
rind (5). During Fe0 oxidation, iron oxide layers of different
oxidation state commonly formed (24). Chemical equilibrium
modeling showed that influent solutions were undersaturated
with respect to CaCO3 minerals, but that pore water solutions
along the columns were supersaturated (Figure 2 in ref 17).
Calcite and aragonite were both detected, but calcite appeared to be abundant only where dissolved Fe was relatively
low (Figure 1b), reflecting the inhibitory effect of ferrous and
ferric ions on calcite precipitation (25, 26).
The fact that columns with different flow rates contain
the same mineral assemblage confirms that the residence
time in the fast flow column (Column I) was sufficient for
major minerals to precipitate. Thus, use of accelerated flow
of high groundwater throughput can simulate mass precipitation resulting from longer periods of “normal” flow, in
this case 17 yrs within the 2-yr study period. However, certain
minerals, such as siderite, although seen in field cores (3-5),
were not detected in column studies here or elsewhere (2).
This may be due to slow kinetics and competition with the
high Ca in groundwater (21, 28). The absence of sulfide
minerals in the present study reflects limited microbial
activity (16). Sulfide formation mediated by sulfate-reducing
bacteria requires a long inoculation time in the presence of

FIGURE 3. TGA-M S results of w eight loss and off gas analysis for Column I core samples (a-c) and stock iron filings (d): (a) from inlet
interface (CS-I-8), (b) from the center of the column (CS-I-81), and (c) from the effluent end of the column (CS-I-49). Note: for sample in
(b), H2 w as purged at 800 instead of 900 °C.
high level SO42- (2). The presence of siderite and iron sulfide
indicates that field PRBs encompass non-steady-state transport and reaction regimes (7, 10, 15), which may not be
adequately simulated in column studies.
Using TGA-MS in Phase Quantification. Weight loss
patterns obtained by TGA-MS are very similar among the
core samples (Figure 3a-c) and are significantly different
from those of stock iron filings (Figure 3d). For all core
samples, the normalized weight decreased gradually to ∼200
°C and then decreased sharply corresponding to the detection
of H2O(g) and CO2(g). Weight loss continued at a slower rate
to 600 °C and then dropped abruptly corresponding to a
sharp increase in CO2. During the isothermal reduction period
at 900 °C, a weight drop accompanied by H2O production
was observed; the weight approached an asymptotic value
(80-91% of initial weight) at the end of the period. For CSI-8, the inlet core sample, a spike of CO2 in addition to H2O
was seen in the off gas (Figure 3a). The source of this CO2
peak is unclear but may be the remaining CO2 from CaCO3
transformation at 600 °C, which was flushed out by H2O.
The TGA-MS results for the stock iron filing (Figure 3d)
show a slight increase in weight at >400 °C, which is attributed
to the oxidation of Fe0 with the surface-absorbed oxygen.
The sample weight started to decrease at >600 °C, with the
occurrence of CO2 in the off-gas. Because the stock iron
contained a small amount of elemental carbon, heating could
enhance C reacting with iron oxide to form Fe0 and CO2 (24).
TGA-MS Analysis with XRD Identification. XRD analysis
was performed on samples that had been heated to 200, 300,

400, 600, 800, and 900 °C (Figure 4). No obvious phase change
occurred at 200 °C. The MS identified the presence of mainly
atomic mass 18 and 17 in the off-gas, representing absorbed
water (23) and its ionized product. A small quantity of CO2
was also detected. The goethite signal disappeared when
samples were heated to 300 °C, but a small broad peak
appeared at 2.7 Å, coinciding with the strong line of hematite
and aragonite. We attribute it to hematite because it persisted
at 600 °C. From 600 to 800 °C, CaCO3 transformed to CaO.
The major phase at 800 °C was Fe3O4, with minor components
of CaO‚Fe2O3, CaO, and Fe2O3 (Figure 4). At 900 °C, FeO,
CaO‚Fe2O3, and CaO were identified. At the end of the
isothermal reduction, the two major phases remaining were
CaO and metallic iron.
The phase changes seen in the XRD analyses are consistent with existing high-temperature phase transformation
data for carbonate and iron-containing minerals. For example, aragonite undergoes polymorphic transition to calcite
at about 450 °C (21). At ∼700 °C, calcite loses CO2 and
transforms into CaO (29). Based on stoichiometry, the CO2
mass can be used to quantify the original mass of CaCO3 in
the sample.
The transformation of iron-containing phases is more
complex than that of CaCO3. Goethite usually loses its
structural OH between 250 and 400 °C (23), according to
2FeOOH f Fe2O3 + H2O. The transformation of Fe2(OH)2CO3 under Ar atmosphere at 500 °C is given by (30)

3Fe2(OH)2CO3 f 2Fe3O4 + H2O + 2H2 + 3CO2
VOL. 38, NO. 21, 2004 / ENVIRONM ENTAL SCIENCE & TECHNOLOGY

9

5761

FIGURE 4. Phase identification of samples (mixtures of CS-I-8, -81, and -49) treated at different temperatures. No obvious phase transformation
betw een the original sample and the sample heated to 200 °C. Goethite signal disappeared betw een temperature of 200 and 300 °C. From
600 to 800 °C, CaCO3 transformed to CaO. The major phase at 800 °C w as Fe 3O4, and, at much lesser intensities, CaO‚Fe 2O3, CaO, and Fe 2O3.
At 900 °C, FeO, CaO‚Fe 2O3, and CaO w ere identified. After heating at 900 °C (w ith 2% H2 in helium), the tw o major phases remaining w ere
CaO and Fe 0, the later resulting from total reduction of iron oxide to metallic iron.
TABLE 2. Precipitate Mass and Porosity Loss Derived from TGA-MS Data (Figure 3)
precipitates (mmol)
samples
stock iron
colum n I
CS-I-8
CS-I-81
CS-I-49
colum n II
CS-II-1
CS-II-3
CS-II-15

as r-FeOOHb
(200-300 °C)

as CaCO3
(600-800 °C)

as Fe 3O4c
(800-900 °C)

initial Fe 0 d
(mmol)

0.024

1.75

precipitate (%)a
CaCO3 + Fe 3O4 + r-FeOOH

corrosion ratese
(mmol kg-1 d-1)

porosity
lossf (%)

0.177
0.223
0.060

0.069
0.167
0.045

0.089
0.097
0.057

1.45
1.25
1.57

29.60
41.71
18.31

4.86
6.16
2.86

45.08
77.57
25.57

0.127
0.024
0.020

0.072
0.017
0.020

0.078
0.032
0.011

1.48
1.66
1.68

32.60
10.16
5.07

4.09
1.50
0.50

40.57
12.10
7.39

a Total m ass percentage of all precipitates corrected for the initially oxidized m ass based on data from the reduction of stock iron filings. b The
m ass of R-FeOOH calculated using m ass of H2O during 200-300 °C heating assum ing goethite is transform ed to hem atite (see Figure 4). c M ass
of total oxidized iron transform ed in this tem perature range, calculated as Fe3O4. d Calculated from the final w eight, less the m ass of CaO. e Calculated
from the m ass ratio of total oxidized iron to the initial Fe0, averaging over the experim ental run tim e (680 d for Colum n I, and 695 d for Colum n
II). For Colum n I the run tim e is equivalent to ∼17-yr of sim ulated norm al flow in Colum n II. f Porosity loss calculated based on 60% initial porosity,
corrected for porosity gain from loss of iron due to corrosion using density values of 2.93 for CaCO3, 5.1 for Fe3O4, 3.8 for R-FeOOH, and 7.9 for
Fe0.

No information is available on the thermo-induced phase
transformation of amorphous iron oxide and green rust. Fe2O3
and Fe3O4 can be quantified from the loss of O atoms during
the isothermal reduction at 900 °C by H2 as follows (24):

3Fe2O3 + H2 f 2Fe3O4 + H2O
Fe3O4 + H2 f 3FeO + H2O
FeO + H2 f Fe0 + H2O
Since reduction of Fe3O4 to FeO occurs above 800 °C,
oxidized iron should be calculated based on the mass
difference at 800 °C and at the end of heating. Although losses
from the oxidation of carbon in iron may occur, the
percentage is small (