ARTICLE IN PRESS

Available at www.sciencedirect.com

WAT E R R E S E A R C H 4 0 ( 2 0 0 6 ) 2 3 1 1 – 2 3 2 0

0043-1354/$ - see frodoi:10.1016/j.watres

�Corresponding a+1 203 432 3068; fax:

E-mail address: y

journal homepage: www.elsevier.com/locate/watres

Profiling iron corrosion coating on iron grains in azerovalent iron system under the influence of dissolvedoxygen

Tian C. Zhang, Yong H. Huang�

Department of Civil Engineering, University of Nebraska-Lincoln, Omaha, NE 68182-0178, USA

a r t i c l e i n f o

Article history:

Received 15 July 2005

Received in revised form

12 April 2006

Accepted 21 April 2006

Keywords:

Zerovalent iron

Iron aquatic corrosion

Iron oxides

Sonication

Dissolved oxygen

nt matter & 2006 Elsevie.2006.04.026

uthor. Present address+1 203 432 5023.

ongheng.huang@yale.edu

A B S T R A C T

Rapid oxidation of Fe0 by O2 occurred when Fe0 grains were bathed in 0.54 mM FeCl2solution saturated with dissolved oxygen (DO), forming a substantial corrosion coating on

Fe0 grains. A sonication method was developed to strip the corrosion coating off the iron

grains layer by layer. The transformation of the constituents and the morphology of the

corrosion coating along its depth and over reaction time were investigated with

composition analysis, X-ray diffraction and scanning electron microscopy. Results indicate

that the sonication method could consistently recover 490% iron oxides produced by the

Fe0-DO redox reaction. Magnetite (Fe3O4) and lepidocrocite (g-FeOOH) were identified as the

corrosion products. Initially, lepidocrocite was the preferential product in the presence of

DO. As the oxide coating thickened, the inner layer transformed to magnetite, which

retained as the only stable corrosion product once DO was depleted. The study confirms the

phase transformations between g-FeOOH and Fe3O4 within a stratified corrosion coating.

The sonication technique exemplifies a new approach for investigating more complicated

processes in Fe0/oxides/contaminants systems.

& 2006 Elsevier Ltd. All rights reserved.

1. Introduction

Iron corrosion is known to be one of the most significant yet

undesirable spontaneous electrochemical processes. For dec-

ades, corrosion prevention has been one of the concerns in

corrosion sciences. In environmental engineering and

science, however, using zerovalent iron (ZVI or Fe0) promoted

processes for in-situ or on-site remediation of environmental

contaminated sites is becoming more and more popular over

the last decade (Klausen et al., 1995; Hansen et al., 1996; US

Army Corps of Engineers, 1997; Scherer et al., 1998, 2000). In

the processes, contaminants are destructed by redox reaction

either with Fe0 or with Fe0-derived reductants like various

Fe(II)-containing compounds or even atomic hydrogen. Con-

r Ltd. All rights reserved.

: School of Environment

(Y.H. Huang).

taminants may also be immobilized by adsorption onto the

iron corrosion products. The application of iron corrosion as

an environmental treatment process requires understanding

of corrosion from a new perspective: in addition to under-

stand the issues such as the effects of pH, dissolved oxygen

(DO), and various corrosive anions on iron aquatic corrosion,

environmental scientists need to develop theories and

technologies to promote the reactivity (or overcome the

passivation) of iron media under the influence of various

contaminants and for long-term remediation purposes.

In iron aquatic corrosion, a stratified profile of corrosion

coating, with inner layer of magnetite and outer layer of

maghemite was reported (Nagayama and Cohen, 1962; Cahan

and Chen, 1982; Oblonsky et al., 1997). In ZVI research,

al Studies, Yale University, New Haven, CT 06511, USA. Tel.:

ARTICLE IN PRESS

WAT E R R E S E A R C H 4 0 ( 2 0 0 6 ) 2 3 1 1 – 2 3 2 02312

progress has been slow in terms of understanding the

character and roles of iron corrosion coating. Recently, spatial

distribution and evolution of iron corrosion products or

precipitates in ZVI systems (e.g., permeable iron barriers)

have been studied from geomineralogical and biomineralogi-

cal perspectives (Gu et al., 1999; Phillips et al., 2000, 2003a, b;

Morrison 2003). It has been found that Fe0 reactivity differs in

part depending on the resulting iron oxides; oxides such as

magnetite or green rusts may promote the reactivity of Fe0,

while others such as maghemite may demote the Fe0

reactivity (Huang et al., 2003; Satapanajaru et al., 2004; Mishra

and Farrell, 2005). An in-situ Raman spectroscopy analysis on

iron surface film (Ritter et al., 2003) indicated that iron

corrosion coating, in particular the surface layer, may evolve

over time. Despite the wide belief that a stratified iron coating

may exist and likely determine the overall reactivity of iron

system, the theory has been supported only by circumstantial

evidence; direct and unmistaken profiling of the corrosion

coatings in a ZVI system is still missing. Studies are still

needed to elucidate the formation, structure, and function of

iron oxide coatings in ZVI processes. Why, so far, only very

limited efforts have been made, with limited progress in this

area, even though a variety of instruments such as X-ray

diffraction (XRD), scanning electron microscopy (SEM), Ra-

man and Mossbauer spectroscopy are available for character-

izing the iron oxides? We believe that this in part reflects the

lack of a proper method for preparing iron oxide samples for

detailed examination of the micro-scale properties of the

corrosion coatings on iron grains.

The objective of the present study was to develop a method

that can be used to profile the iron oxides coated onto iron

grains in a ZVI system. Iron corrosion caused by DO

accelerated by the presence of aq. Fe(II) (Huang and Zhang,

2005) was employed to develop the method. This paper

presents a sonication technique capable of stripping the iron

oxide coating off iron grains layer by layer. By using the

sonication method, we captured the evolving spectroscopic

profiles of iron corrosion coatings as a function of reaction

time and the depth. The results are interpreted and the

implications of our methods are also briefly discussed.

2. Materials and methods

2.1. Materials and chemicals

Unless otherwise indicated, all aqueous solutions were

prepared using deionized water (DI) with resistivity between

15 and 18 MO-cm (Barnstead Nanopure series 550, Barnstead/

Thermolyne Co., Dubuque, IA). The deoxygenated water (or

solution) was made by flushing DI water (or solution) with

analytical-grade argon and stored in an argon environment

before use. Argon (99.999%) and oxygen (99.99%) gas were

provided by Linweld, Inc. (Lincoln, NE). All commercially

available chemicals and minerals were used as received.

Ferrous iron (Fe2+) was prepared from FeCl2 � 4H2O (J.T. Baker

Co., Phillipsburg, NJ). The industrial iron grains were largely

free from visible rust and retained a metallic glaze (US Metals

Co., Chicago, IL). The iron particles were approximately 0.5-

mm in diameter, irregular in shape, with a slightly rough

surface and a BET (Brunauer-Emmett-Teller) surface area of

0.04 m2 g�1. The iron grains were used without further pre-

treatment.

2.2. Experimental methods

All batch tests employed serum bottles (VWR Scientific, IL)

capped with rubber stoppers as reactors. A series of batch

tests were conducted, all with initial conditions of 0.5 g Fe0

grains+10 mL of solution containing 30 mg Fe2+ L�1+ 4.05 mL

1 atm O2 in headspace. The initial conditions was selected for

creating the iron corrosion coating because: (1) the reaction

under the test condition is not too slow, thus allowing

numerous repeats to produce samples; (2) since processing

a sample may take up to 25 min, the time lag may introduce

considerable error if a faster reaction is involved, and (3) the

selected test exhibited most distinct reaction stages, allowing

us to capture the transformation of iron oxide coating as well

as producing more distinguishable profiles. The details of the

reaction mechanism were reported in our previous study

(Huang and Zhang, 2005).

In each test, multiple reactors were prepared using the

following procedures: (1) Stock solutions (i.e., 30 mg Fe2+ L�1

as FeCl2) were prepared by dissolving pre-weighted chemicals

in DI water and then deoxygenated before use, (2) ten

milliliters of the stock solution was transferred to the reactor

containing pre-weighted iron grains, (3) the reactor was

immediately capped with a stopper, the headspace

(4.0570.06 mL) was flushed with O2 for 20 s by inserting two

needles through the stopper and allowed the headspace

pressure to equilibrate with the atmosphere so that the initial

O2 pressure in the headspace was 1.0 atm, (4) the reactors

were placed in a 30�45 cm box rotating at 30 rpm to provide

complete mixing in the dark, and (5) at selected times, one

reactor was sacrificed for regular analyses of various para-

meters. In these experiments, only the initial conditions in

the reactors were controlled.

2.3. Sonication and sample preparations

All tests and analyses were conducted at room temperature

(2371 1C). Whenever possible, samples were prepared in an

anaerobic box that was filled with argon gas. For regular

analyses, the serum bottles (n ¼ 3) taken from the rotator was

immediately transferred into the anaerobic box. For O2

consumption, the headspace pressure was measured with a

25 mL gas tight syringe as per the method described before

(Huang and Zhang, 2005). The reactor was then opened, and

5 mL solution (together with precipitates if presented but

avoiding iron grains) was withdrawn with a syringe. The pH of

the reactor solution was immediately measured. The 5 mL in

the syringe was then filtered through 0.45-mm membrane disc

filter (Pall Gelman). The filtrate was analyzed for dissolved

Fe2+, Fe3+, and Cl�.

For profiling corrosion coatings with the sonication techni-

que, separate reactors (n ¼ 2 or 3) were used with the

following procedures. (1) After transferring the reactors into

the anaerobic box, the 10 mL solution (without iron grains but

with precipitates if existed) was withdrawn with a syringe as

thoroughly as possible and filtered through a filter (0.45-mm,

ARTICLE IN PRESS

0

20

40

60

80

100

0 4 8 12 16 20 24Time (h)

0 4 8 12 16 20 24Time (h)

O2

Rem

oval

(%

)

0.0

5.0

10.0

15.0

20.0

25.0

Mas

s of

Pro

duct

s (m

g)

0

5

10

15

20

25

30

35

Dis

solv

ed F

e2+ (

mg

L-1)

6.0

5.0

4.0

3.0

2.0

1.0

0.0

pH

O2

Fe3O4

FeOOH

Fe2+

pH

(a)

(b)

Fig. 1 – Oxygen removal in a ZVI system augmented with

30 mg Fe2+ L�1. (a) Percentage of oxygen consumption

(measured data) and the accumulated mass of Fe3O4 or

FeOOH (calculated based on Eqs. (1)–(3)), and (b) change of

dissolved Fe2+ and pH over time.

WAT E R R E S E A R C H 40 (2006) 2311– 2320 2313

Gelman). The precipitate would be retained as an oxide film

on the filter paper, which was dried in a capped jar with argon

flushing. Once leaving the anaerobic box, the sample was

further dried in a desiccator for 24 h at ambient temperature

before weighing. The weight difference of the filter paper

before and after filtration was considered as the mass of the

precipitate. The samples were saved for XRD characterization

and for analysis of the molar ratio of Fe(II) and Fe(III)

constituents (see below). (2) The reactor (with iron grains)

was then re-filled with 10 mL DI water, capped and then put in

an ultrasonic cleaner (1.9 L, 150 W, VWR Scientific) for

sonication. The reactor was fixed with a positioner on the

center of the cleaner to ensure an identical sonication

condition for all reactors. Once the sonication was turned

on, the oxide coatings would start to be stripped off and

become precipitate. After sonicating for a specific time (e.g.,

1 min), the 10 mL solution (without iron grains but with the

precipitates) was thoroughly withdrawn and then filtered.

The retained oxide on the filter paper represented the outer

layer of the corrosion coating. Similarly, the sample was

dried, weighed, and then saved. (3) Once the outer layer was

removed, multiple (e.g., 2–3 times) sonications of certain

durations could be repeated by following step (2) to retrieve

the middle and then the inner layer of the oxide coating. After

numerous trials, we concluded that an added-up sonication

duration of 10-min would be sufficient to strip off most of the

retrievable oxide coatings from iron grains. When a multiple-

sonication operation was employed, the sonication time

could be distributed into (a) 1 min for the outer layer+1 min

for the middle layer+8 min for the inner layer, sequentially; or

(b) 1 min for the outer layer+9 min for the inner layer

sequentially when the coating was not thick enough. (4)

Parallel reactors at any of steps 1–3 may be sacrificed to

prepare SEM samples. The iron grains before or after certain

sonication treatment were rinsed with DI water for 3 times,

then dried with Ar flushing for �20 min; and then saved in an

argon environment until SEM analyses with the storage time

shorter than 24 h. The micrographs on the corrosion coatings

of different depths exposed after sonication treatments

generate a morphological profile of the oxide coating.

To obtain the molar ratio of Fe(II) to Fe(III) of iron oxide

samples, the filter papers together with the retained oxide

film were put into a sealed serum bottle filled with Ar gas, and

then injected 1 mL 6 N HCl with a syringe to soak the oxide

sample. After 20 min when the iron oxides would be

completely dissolved, the solution was then diluted by

10–50 times with deoxygentated DI water for HPLC analysis

of Fe2+ and Fe3+. In a quality control test with commercial

magnetite (FeIIFeIII2 O4), the same procedure produced a valid

Fe(II)/Fe(III) molar ratio of about 1:2.0, indicating our proce-

dure was sound. Exposure of aqueous Fe(II) to O2 under strong

acidic condition (e.g., 6 N HCl, but not the diluted sample)

should be avoided as our control showed that Fe2+ may be

slowly oxidized to Fe3+.

2.4. Analytical methods

The pH of the reactor was measured with an ORION semi-

micro pH probe. A Dionex DX 500 HPLC/IC (high performance

liquid chromatography/ion chromatography) system (Dionex

Co., Sunnyvale, CA) was used to analyze Fe2+, Fe3+ and Cl�

ions involved in the system as per procedures reported before

(Huang et al., 2003). The XRD was equipped with a peltier-

cooled, solid-state detector (PAD V model, Scintag Inc.,

Cupertino, CA). The operating wavelength was 1.5405mm (Cu

source). The library used for spectrum matching was from the

JCPDS-International Center for Diffraction Data (PCPDFWIN v.

2.01, 1998). Scanning electron microscopy (SEM. HITACHI

S-4700 N, Tokyo, Japan) was used to discern the features of the

source iron, the oxide coating(s), and the precipitate(s) formed

under the test conditions.

3. Results

As shown in Fig. 1, 495% of oxygen would be consumed

within 6 h (Fig. 1a). Initially, a small amount of dissolved Fe2+

(o7 mg L�1) was used accompanied with a pH increase from

4.4 to 5.2. Both pH and dissolved Fe2+ were stable after 6 h

reaction. Brownish coating on iron grains was visible in the

first 3 h. Between t ¼ 224 h, brownish precipitate started to

form and accumulated. By t ¼ 4 h, near 88% oxygen had been

removed, the brownish oxide coating turned black and the

brownish precipitate receded to black precipitate. After t ¼ 6 h

when 495% oxygen removed, both the coating and the

precipitate were completely black in color.

ARTICLE IN PRESS

WAT E R R E S E A R C H 4 0 ( 2 0 0 6 ) 2 3 1 1 – 2 3 2 02314

As shown in Table 1, the actual mass recovery of the

sonicated oxide coating increases with time, so is the

theoretical mass of magnetite or lepidocrocite formed

due to the oxygen consumption. The composition (FeII/FeT)

of the oxide coating also changes with time and depth.

Therefore, the mass and constituent distributions of the

corrosion coatings are a function of both reaction time and

the depth inside the coating. These distributions also indicate

that the iron oxide coating can be stripped off layer by layer

with the sonication technique.

Some XRD spectra of samples from the inner layer show a

slightly bending baseline before 251 of 2y. However, the well-

formed peaks of the XRD spectra clearly indicate that all iron

corrosion products formed from the very beginning and at

different depths were well crystalline in structure (Fig. 2). In

contrast, iron oxides formed in a ZVI system operated in an

acidic environment for nitrate reduction (Huang and Zhang,

2004) are mostly amorphous and would subsequently take �1 d

to age into a crystalline structure. As shown in Fig. 2,

lepidocrocite and magnetite are identified as the two major

corrosion products of Fe0–O2 reaction. Overall, lepidocrocite was

the dominant product until 4 h. After 6 h, when most DO was

consumed, magnetite became the dominant species (Fig. 2).

The most representative SEM micrographs from each

sample, after examining more than a dozen grains on various

surface spots for each sample, are presented in Fig. 3. The

micrographs show that the sonication process would not

rupture the basic structure of the iron grains. No significant

iron oxide exists in the original iron grains [Fig. 3(a)]. After

reacting 1 h, significant amount of lepidocrocite had amassed

on the surface [Fig. 3(b) and (c)]. Fig. 3(b) shows an

undisturbed vertical profile of the corrosion coating. The

lepidocrocite has a flake- or scale-like structure, which is

agreeable with literature report (Cornell and Schwertmann,

1996). Apparently, the structure was porous with a large

surface area. However, when the surface layer was peeled off

after 1 min sonication, it exposed the underlying structure of

granular aggregates [Fig. 3(d)]. Most of the granular structure

seemed to be in its early forming stage. The structure

compacts more tightly, leaving no or very limited pore in

the inner layer. After 10 min sonication, the surface of

the iron grain was almost as clean as the original iron grain

[Fig. 3(e)], confirming that most of the iron corrosion coating

had been removed. This result is consistent with the 94.5%

mass recovery of oxide coating stripped off after 10 min

sonication (Table 1).

The morphology of the iron corrosion coating in the

samples after 2 or 4 h reaction was not much different from

the sample of 1 h, only with more granular feature developed

from the flake-like structure in the outer layer. Fig. 3(f–h)

shows the morphological transformation of lepidocrocite to

magnetite which characterize the samples from 4 h reaction.

In some cases, it seems like a tightly aggregated granular

structure was evolved from the flake structure. However, the

surface after 10 min sonication still retained some iron oxides

[Fig. 3(h)], indicating that the iron oxides adhered more tightly

to the underlying Fe0. On the samples from t ¼ 24 h, only

granular magnetite crystalline remained.

Since both SEM and XRD are conducted on dried samples,

potential dehydration of the oxide samples may complicate

interpretation of the results. For instance, the iron oxide

formed by Fe0 nitrate reaction at pH 2.5 was found to be

evolving during a similar drying procedure (Huang and Zhang,

2004). To dispel the concern, a control test was conducted to

show that XRD results on the wet samples were not

essentially different from the parallel samples after drying

procedure (data not shown). For the former, the fresh wet

sample was covered with cellophane tape to insulate from

oxygen exposure and scanned with XRD within minutes. In

fact, magnetite and lepidocrocite are not known to transform

rapidly at ambient temperature and circumneutral pHs

involved in this study. The other concern is that XRD may

not differentiate magnetite from maghemite. Fortunately, the

Fe(III)/Fe(II) molar ratios close to 2:1 indicate that magnetite

was the main component. Nonetheless, the magnetite we

refer hereto may be more rigorously understood as a spinel

structure with flexible partitions of Fe(II) and Fe(III) occupying

their respective lattices; i.e., a non-stoichiometric ratio of

Fe(III)/Fe(II) larger than 2 that yielded a XRD spectrum we

identified hereto as magnetite could be seen as a mixture of

magnetite and a small amount of maghemite (Jolivet et al.,

2004).

4. Discussion

4.1. Methodology

First, calculation of the theoretical mass of magnetite or

lepidocrocite formed due to oxygen consumption is based on

the following three reactions (Huang and Zhang, 2005):

4 Fe0 þ 3 O2 þ 2 H2O! 4 g�FeOOH (1)

8 g�FeOOHþ Fe0 ! 3 Fe3O4 þ 4 H2O (2)

4 Fe3O4 þO2 þ 6 H2O! 12 g�FeOOH (3)

Eq. (1) occurs at the very beginning of the reaction. Once the

iron surface is covered by g-FeOOH and oxygen diffusion is

not fast enough, Fe0 will react with g-FeOOH in the inner layer

of iron corrosion coating (Eq. (2)). Our XRD and SEM results

seem to support Eqs. (1) and (2). Eq. (3) occurs in the

oxide–liquid interface when DO is available. It should be

noted that in the presence of DO, both Eqs. (2) and (3) occur

and which one dominates depends on the dynamics and

oxygen diffusion into the inner layer of the oxide coating.

Once the oxic condition ceases, Eq. (3) stops while Eq. (2)

continues until magnetite become the sole product of the

system. These inferences also are supported by our SEM and

XRD evidence. The role of Fe2+ in facilitating the electron

transfer through the oxide coating, though still murky at this

stage, may be better understood as a ‘‘chemi-conductor’’

mechanism proposed by Cahan and Chen (1982).

Other than through an autoreduction reaction represented

by Eq. (2), transformation of g-FeOOH to Fe3O4 may also occur

in the presence of surface-adsorbed Fe(II) (Tamaura et al.,

1981):

2 g�FeOOHþ Fe2þ ! Fe3O4 þ 2 Hþ (4)

Eq. (4) occurs at ambient temperature, but only with a

system pH47.3, a condition required for the adsorption of

ARTICLE IN PRESS

Ta

ble

1–

Ma

ssre

cov

ery

an

dco

mp

osi

tio

no

fp

reci

pit

ate

sa

nd

iro

no

xid

eco

ati

ng

stri

pp

ed

off

wit

hso

nic

ati

on

treatm

en

t

RT

1(h

)aR

T2

(h)

RT

4(h

)R

T6

(h)

RT

12

(h)

RT

24

(h)

Sa

mp

leso

urc

eM

ass

(mg)

bFe

II/

FeT

(%)c

Ma

ss(m

g)Fe

II/

FeT

(%)

Ma

ss(m

g)Fe

II/

FeT

(%)

Ma

ss(m

g)Fe

II/

FeT

(%)

Ma

ss(m

g)

FeII/

FeT

(%)

Ma

ss(m

g)Fe

II/

FeT

(%)

Pre

cip

ita

tein

bu

lkso

luti

on

0N

/A0

N/A

3.8

75.8

10.4

712.3

12.2

023.5

%14.3

330.8

%

Ox

ide

coa

tin

gS

trip

ped

off

by

son

ica

tio

nfo

rd

1m

in4.1

04.1

6.4

74.4

9.0

713.8

4.5

719.4

3.4

027.1

%2.4

031.5

%

1m

in1.7

013.0

3.5

315.5

2.2

715.8

1.9

326.2

1.9

334.5

%1.2

336.0

%

8m

in1.4

725.4

1.1

030.1

Act

ua

lm

ass

reco

very

of

son

ica

ted

ox

ide

coa

tin

g

5.8

07

0.1

0e

10.0

07

0.8

716.6

77

1.0

818.0

77

0.7

117.5

37

0.3

217.9

77

0.4

0

O2

up

tak

e(%

)f32%

56%

88%

96%

100%

100%

Th

eo

reti

cal

ma

sso

fFe

3O

4(

org-

FeO

OH

)

ba

sed

on

O2

use

dg

6.1

4(6

.27)

10.7

5(1

0.9

8)

16.9

0(1

7.2

5)

18.4

3(1

8.8

2)

19.2

0(1

9.6

0)

19.2

0(1

9.6

0)

Reco

very

(%)

94.5

%93.0

2%

98.6

2%

96.0

%91.3

%93.6

%

aR

T-R

ea

ctio

nti

me

(h).

bM

ass

rep

ort

ed

isth

eavera

ge

of

trip

lica

tes

(n¼

3)

un

der

init

ial

con

dit

ion

s:0.5

gFe

0gra

ins+

30

mg

Fe2+

L�

1+

4.0

5m

l1

atm

O2.

cFe

II/F

eT¼

the

av

era

ge

of

du

pli

cate

s(n¼

2).

FeII

(or

FeII

I )in

clu

des

sorb

ed

an

dst

ruct

ura

lFe

2+

(or

Fe3+)in

the

pre

cip

ita

teo

rir

on

ox

ides,

FeT¼

FeII+

FeII

I ,wh

ich

iso

bta

ined

by

firs

ttr

ea

tin

gth

ep

reci

pit

ate

or

stri

p-o

ffir

on

ox

ides

att

ach

ed

on

the

filt

er

pa

per

wit

hH

CL

tore

lea

seFe

2+

an

dFe

3+

into

the

solu

tio

n,

an

dth

en

an

aly

zin

gFe

2+

an

dFe

3+

wit

hth

eH

PLC

.d

So

nic

ati

on

can

be

2ti

mes

(1+

9m

in)

or

3ti

mes

(1+

1+

8m

in)

wit

hto

tal

ad

d-u

pti

me

of

10

min

.e

5:8

0¼

4:1

0þ

1:7

0;7

0.1

0is

the

sta

nd

ard

dev

iati

on

.f

resu

lts

fro

mth

eb

atc

hte

st(F

ig.

1a

).g

Ca

lcu

late

db

ase

do

nE

qs.

(1)–

(3)

wit

hm

agn

eti

te(o

rle

pid

ocr

oci

te)

as

the

sole

pro

du

ctfo

rca

lcu

lati

on

.

WAT E R R E S E A R C H 40 (2006) 2311– 2320 2315

ARTICLE IN PRESS

10 15 20 25 30 35 40 45 50 55 60 65 70Degree (2 theta) Degree (2 theta)

10 15 20 25 30 35 40 45 50 55 60 65 70

10 15 20 25 30 35 40 45 50 55 60 65 70Degree (2 theta)

10 15 20 25 30 35 40 45 50 55 60 65 70Degree (2 theta)

Degree (2 theta)10 15 20 25 30 35 40 45 50 55 60 65 70

M

L

1

2

ML

1

2

3

P

M

L

1

2

ML

1

2

3

P

M

P

1

2

(a) (c)

(b)

(e)

(d)

Fig. 2 – Depth profiles of compositions of the iron corrosion coating by XRD over the course of O2–Fe0 reaction. Corrosion

coating from t ¼ (a) 1 h; (b) 2 h; (c) 4 h; (d) 6 h; and (e) 24 h. Spectrum P is from the precipitate before sonication. Spectra 1, 2, 3

represent the samples collected after sequential sonications, following the respective time schemes reported in Table 1 (for

(a), (b) and (e), 1+9 min; for (c) and (d), 1+1+8 min, respectively). Spectra M and L: magnetite and lepidocrocite XRD spectra,

respectively, adopted from the standard library.

WAT E R R E S E A R C H 4 0 ( 2 0 0 6 ) 2 3 1 1 – 2 3 2 02316

Fe2+ onto g-FeOOH to form an intermediate. Based on slight

changed pH and the stable dissolved Fe2+ concentration in our

system (Fig. 1b), it appears that the transformation of

lepidocrocite did not follow Eq. (4). However, Eq. (4) might

occur on the surface coating in a microscopic environment

where a local pH above 7.3 is possible.

It should be pointed out that Eqs. (1)–(3) are not derived

based on mechanisms (or oxidation–reduction half reactions)

involved in our system but on our observations of iron

corrosion products and mass balances. Based on Eqs. (1)–(3),

the theoretical mass of corrosion products from O2–Fe0 redox

reaction could be predicted, i.e., 19.2 mg Fe3O4 or 19.6 mg g-FeOOH produced by 4.05 mL O2 at 24 1C and local barometric

pressure of 9.97�104 Pa (or 0.1656 m-mole O2). As shown in

Fig. 1a, the predicted magnetite production differs little from

that of lepidocrocite; either one, therefore, can be used as the

theoretical mass of iron oxides produced due to oxygen

consumption. Our results indicate that using sonication can

consistently recover 91–98% of the theoretical iron corrosion

products. According to our observation, the loss may be due

to (1) the inner most layer of iron corrosion coating adhering

so tightly to Fe0 that it cannot be removed by sonication, (2)

ARTICLE IN PRESS

Fig. 3 – Morphological profiles of the iron corrosion coating. SEM (20 or 50 k) of (a) original surface of iron grains, (b) vertical

profile of the oxide coating shown as a broken edge, sampled from t ¼ 1 h before sonication, (c) lepidocrocite as the outer

layer, from t ¼ 1 h before sonication, (d) the inner layer, from t ¼ 1 h after 1 min sonication, (e) typical surface after 10 min

sonication, from t ¼ 1 h after 10 min sonication, (f) transformation of lepidocrocite to magnetite, from t ¼ 4 h before

sonication, (g) magnetite crystalline predominant, from t ¼ 4 h after 1 min sonication, and (h) magnetite aggregate of inner

layer, from t ¼ 4 h after 8 min sonication.

WAT E R R E S E A R C H 40 (2006) 2311– 2320 2317

corrosion products formed deep in the crevices or cracks of

the iron grains may not be stripped off, (3) loss of the mass

during the operations, e.g., residuals attached on the wall of

the reactors, the syringe and filter holders. SEM micrographs

confirmed the loss due to (1) and (2). Despite these possible

losses, the mass recovery percentages are acceptable.

ARTICLE IN PRESS

WAT E R R E S E A R C H 4 0 ( 2 0 0 6 ) 2 3 1 1 – 2 3 2 02318

Second, the sonication method established in this study

can be used to profile the iron corrosion coating involved in a

ZVI system. However, the method should not be considered

as an in situ analysis. This is because: (1) we assumed that the

Fe0–O2 reaction would stop immediately after the original

reactant solution in the reactor was replaced with equal

amount of deoxygenated DI water, which is difficult to prove,

(2) we neglected the possible continuing transformation

within the corrosion coating during the sample preparing

period (up to 20 min), and (3) the results, especially the ratio of

FeII/FeT within each layer, should only be interpreted from a

statistical perspective. For instance, the first 1 min sonication

is supposed to strip off the outer layer; however, because of

the non-uniformed nature of iron grain surface, a small part

of the iron oxide coating in the inner layer may be stripped off

and thus be counted in as that in the outer layer. Moreover,

we hope that sonication minimally disturbs the original

structure of the corrosion coating, and thus the SEM micro-

graphs still represent the original structure of the method. It

is not easy to verify this anticipation even though it seems to

be the case by comparing the photos of the samples from pre-

sonication (Fig. 3b) with those of post-sonication.

4.2. Stratified structure and evolution of iron oxidecoating

The XRD and SEM results reveal a stratified feature of the

corrosion coating with an outer layer of lepidocrocite and an

inner layer of magnetite. The results also captured the

dynamic transformation from lepidocrocite, the dominant

product under the condition of abundant O2, to magnetite

when DO was exhausted. More importantly, it has to be noted

that the inner layer of the sample from t ¼ 1 h already

consisted of a significant amount of magnetite, suggesting

that magnetite inner layer existed from the very beginning.

During the reaction period, magnetite propagated outward

and prevailed over lepidocrocite. Finally when DO was

20.0

18.0

16.0

14.0

12.0

10.0

8.0

6.0

4.0

2.0

0.0t=1h t=2h t=4h

Iron Corrosio

Mas

s of

Iron

Oxi

des,

Tot

al o

r

Diff

eren

tiate

d (m

g)

Precipitate

Outer layer

Middle layer

Inner layer

4.1%

13%

4.4%

15%

Fig. 4 – Mass and constituent distributions of iron corrosion pr

sonication (as strip-off) as a function of reaction time and depth

each column is the percentage of Fe(II)/[Fe(III)+Fe(II)] in the corre

shown in Table 1.

depleted, only magnetite remained as the stable corrosion

product under the anaerobic condition. No Fe0 peak was

detected in any of the XRD analyses, indicating that no Fe0

was in the samples.

It is instructive to further evaluate magnetite formation as a

function of reaction time and location inside the iron coating.

While lepidocrocite contains Fe(III) only, magnetite with a

perfect structure and no cation substitution or vacancy has

33.3% of Fe(II) and 66.7% of Fe(III). Thus, the FeII/FeT value in

Table 1 provides us the approximate percentage of iron that is

in a magnetite form; e.g., 15% FeII/FeT means that the sample

consisted of about 45% of magnetite (1FeII:2FeIII¼

15%FeII:X%FeIII, X% ¼ 30%, so iron in magnetite form ¼

15%+30% ¼ 45%).

Fig. 4 illustrates magnetite formation vs. reaction time and

oxide locations. At t ¼ 1 h, the outer layer of the oxide coating

(stripped off with 1 min sonication) contained4�12% magne-

tite, while the inner layer (stripped off with 9 min sonication)

contained �39% magnetite. At t ¼ 4 h, the outer layer of the

oxide coating contained �41% magnetite, while the middle

and inner layer (stripped off in sequence with 1 and 8 min

sonication) contained �47% and 76% magnetite, respectively.

It must be noted that the estimate of magnetite distribution

may not be accurate, because the Fe(II)/Fe(III) ratio of the

magnetite may vary in a noticeable range because of

magnetite’s nonstoichiometry nature (represented by Fe3�xO4,

0pxp0.1) (Cornell and Schwertmann, 1996). In addition,

while a distinct Fe(II) profile over depth is attainable, the

percentage of Fe(II) within each layer needs to be interpreted

from a statistical perspective.

Figs. 2–4 are consistent with each other and, together,

provide a clear picture on evolution of the corrosion coating

with time and the constituent distribution in space (i.e., along

the depth of the oxide coating). Collated Table 1, Fig. 4, SEM

micrographs with the XRD results, the spherical crystalline

(�0.1–0.15mm) should be magnetite and the flake-structure,

lepidocrocite. The SEM profiles indicate that magnetite was

t=6h t=12 t=24hn Coating at Time t

5.8%

13%

16%

25%

30%

32%

12%

19%

23%

27%

35% 36%

26%

30%

oducts retrieved before sonication (as precipitates) and after

in oxide coating. The percentage reported on the right side of

sponding samples. Data and the related information are also

ARTICLE IN PRESS

WAT E R R E S E A R C H 40 (2006) 2311– 2320 2319

formed on the iron surface. Therefore, lepidocrocite might be

formed only after the magnetite was further oxidized by the

DO in the solution. Whether the transformation between

lepidocrocite and magnetite is through a solid state reaction

or through a dissolution-crystallization process or through

both (Jolivet et al., 2004) is difficult to determine. Ionization of

Fe0 and migration of Fe(II) or Fe(III) must be involved since

most if not all of Fe(II) and Fe(III) are sourced from the

underlying Fe0.

4.3. Implications

A stratified profile of corrosion coating, with an inner

magnetite layer coupled with an outer layer of more

complicated compositions with higher iron oxidation states,

may exist widely in a more complicated system (Nagayama

and Cohen, 1962; Oblonsky et al., 1997). Since the nature of

the iron oxide coatings on Fe0 is critical to affect the extent of

ZVI reactivity and corrosion, understanding the formation,

structure, and constituent distribution of iron oxides are

important to promote the applications of ZVI processes for

remediation. Profiling the iron corrosion coatings may help

identify the reactive site, evaluate reaction dynamics, locate

effective adsorptive components, and determine the factors

that are responsible for ZVI passivation.

In this study, we developed a sonication technique that can

strip the iron oxide coating off iron grains layer by layer with

�90% mass recovery. The method can have many applica-

tions in either ZVI systems or surface-mediated processes; it

can serve as a platform for further improvement of the

methodology as well.

5. Conclusions

Fe0 grains were bathed in 0.54 mM aqueous Fe2+ and saturated

dissolved oxygen (DO), allowing rapid redox reaction between

Fe0 and DO that resulted in an iron oxide coating on the Fe0

grains. A sonication procedure was demonstrated capable of

stripping off the iron corrosion coating layer by layer,

recovering over 90% of the iron oxides. Spectroscopic profiles

of the corrosion coating indicate that a stratified structure,

characterized by a porous lepidocrocite outer layer and a

more compact magnetite inner layer, was formed. The

distribution of Fe(II) richness in the oxide coating over depth

is consistent with a ‘‘chemi-conductor’’ model proposed by

Cahan and Chen (1982), and may explain the role of added

aqueous Fe2+ in terms of promoting the reactivity of a

zerovalent iron system under the test conditions. It may also

explain the subsequent autoreduction of lepidocrocite to

magnetite by the underlying Fe0 upon depletion of the DO in

the system.

Acknowledgments

The authors gratefully acknowledge Dr. P.J. Shea and Dr. S.D.

Comfort, School of Natural Resource Sciences, Univ. of

Nebraska-Lincoln (UNL) for their important suggestions and

comments during the project, Mr. B.E. Johns, Center for

Materials Research and Analysis at UNL for XRD analysis,

Dr. K.J. Klabunde, Dept. of Chemistry, Kansas State University,

for BET surface area analysis and Dr. K. Lee, School of

Biological Sciences, UNL for SEM analysis. This research

was supported in part by the US EPA/EPSCoR Program

(Proj. R-829422-010) and the Nebraska Research Initiative.

The College of Engineering and Technology at UNL provided

matching funds for the project.

R E F E R E N C E S

Cahan, B.D., Chen, C.T., 1982. The nature of the passive film oniron. J. Electrochem. Soc. 129, 921–925.

Cornell, R.M., Schwertmann, U., 1996. The Iron Oxides: Structure,Properties, Reactions, Occurrence and Uses. VCH Publishers,New York.

Gu, B., Phelps, T.J., Liang, L., Dickey, M.J., Roh, Y., Kinsall, B.L.,Palumbo, A.V., Jacobs, G.K., 1999. Biogeochemical dynamics inzero-valent iron columns: implication for permeable reactivebarriers. Environ. Sci. Technol. 33, 2170–2177.

Hansen, H.C.B., Koch, C.B., Nancke-Krogh, H., Borggaard, O.K.,Sorensen, J., 1996. Abiotic nitrate reduction to ammonium: keyrole of green rust. Environ. Sci. Technol. 30, 2053–2056.

Huang, Y.H., Zhang, T.C., 2004. Effect of low pH on nitratereduction by iron powder. Water Res. 38, 2631–2642.

Huang, Y.H., Zhang, T.C., 2005. Effects of dissolved oxygen onformation of corrosion products and concomitant oxygen andnitrate reduction in zero-valent iron systems with or withoutaqueous Fe(II). Water Res. 39, 1751–1760.

Huang, Y.H., Zhang, T.C., Shea, P.J., Comfort, S.D., 2003. Effect ofiron coating and selected cations on nitrate reduction by iron.J. Environ. Qual. 32, 1306–1315.

Jolivet, J.P., Chaneac, C., Tronc, E., 2004. Iron oxide chemistry: frommolecular clusters to extended solid networks. Chem. Com-mun. 2004, 481–487.

Klausen, J., Trober, S.P., Haderlein, S.B., Schwarzenbach, R.P., 1995.Reduction of substituted nitrobenzenes by Fe(II) in aqueousmineral suspensions. Environ. Sci. Technol. 29, 2396–2404.

Mishra, D., Farrell, J., 2005. Understanding nitrate reactions withzerovalent iron using Tafel analysis and electrochemicalimpedance spectroscopy. Environ. Sci. Technol. 39, 645–650.

Morrison, S., 2003. Performance evaluation of a permeablereactive barrier using reaction products as tracers. Environ.Sci. Technol. 37, 2302–2309.

Nagayama, M., Cohen, M., 1962. The anodic oxidation of iron in aneutral solution: 1. the nature and composition of the passivefilm. J. Electrochem. Soc. 109, 781–790.

Oblonsky, L.J., Davenport, A.J., Ryan, M.P., Isaacs, H.S., Newman,R.C., 1997. In-situ X-ray absorption near edge structure studyof the potential dependence of the formation of the passivefilm on iron in borate buffer. J. Electrochem. Soc. 144,2398–2404.

Phillips, D.H., Gu, B., Watson, D.B., Roh, Y., Liang, L., Lee, S.Y., 2000.Performance evaluation of a zerovalent iron reactive barrier:mineralogical characteristics. Environ. Sci. Technol. 34,4169–4176.

Phillips, D.H., Gu, B., Watson, D.B., Roh, Y., 2003a. Impact ofsample preparation on mineralogical analysis of zero-valentiron reactive barrier materials. J. Environ. Qual. 32, 1299–1305.

Phillips, D.H., Watson, D.B., Roh, Y., Gu, B., 2003b. Mineralogicalcharacteristics and transformations during long-term opera-tion of a zerovalent iron reactive barrier. J. Environ. Qual. 32,2033–2045.

Ritter, K., Odziemkowski, M.S., Simpgraga, R., Gillham, R.W., Irish,D.E., 2003. An in situ study of the effect of nitrate on the

ARTICLE IN PRESS

WAT E R R E S E A R C H 4 0 ( 2 0 0 6 ) 2 3 1 1 – 2 3 2 02320

reduction of trichloroethylene by granular iron. J. Contam.Hydrol. 65, 121–136.

Satapanajaru, T., Shea, P.J., Comfort, S.D., Roh, Y., 2004. Green rustand iron oxide formation influences metolachlor dechlorina-tion during zerovalent iron treatment. Environ. Sci. Technol.37, 5219–5227.

Scherer, M.M., Balko, B.A., Tratnyek, P.G., 1998. Mineral–waterinterfacial reactions: kinetics and mechanisms. In: Sparks,D.L., Grundl, T.J. (Eds.), ACS Symposium Series 715.American Chemical Society, Washington, DC,pp. 301–322.

Scherer, M.M., Richter, S., Valentine, R.L., Alvarez, P.J., 2000.Chemistry and microbiology of permeable reactive barriersfor in situ groundwater cleanup. Crit. Rev. Environ. Sci.Technol. 30, 363–411.

Tamaura, Y., Buduan, P.V., Katsura, T., 1981. Studies on theoxidation of iron(II) ion during the formation of Fe3O4 and a-FeOOH by air oxidation of Fe(OH)2 suspensions. J. Chem. Soc.Dalton Trans. 1981, 1807–1811.

US Army Corps of Engineers, 1997. Design Guidance for Applica-tion of Permeable Barriers to Remediate Dissolved ChlorinatedSolvents.