PII: S0043-1354(00)00591-1

Wat. Res. Vol. 35, No. 12, pp. 29612969, 2001

# 2001 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0043-1354/01/$ - see front matter

PHYSICO-CHEMICAL CHARACTERISTICS OF CORROSION

SCALES IN OLD IRON PIPES
P. SARIN2, V. L. SNOEYINK1*, J. BEBEE1, W. M. KRIVEN2 and J. A. CLEMENT3
1

Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign,

IL 61801, USA; 2 Department of Materials Science and Engineering, University of Illinois at
Urbana-Champaign, IL 61801, USA and 3 Black and Veatch, Cambridge, MA 02140, USA
(First received 1 July 2000; accepted in revised form 1 December 2000)

Abstract}Corrosion scales play an important role in modifying water quality in drinking water
distribution systems. The corrosion scales from old iron/steel pipes were analyzed for their structure and
composition. Scales were studied both before and after drying, and goethite, magnetite and lepidocrocite
were identied as the primary constituents of the dried samples. High concentrations of readily soluble
ferrous phases were detected in wet-scale samples. The corrosion scales had a shell-like, enveloping layer,
covering porous deposits of iron oxide phases. Our studies were able to identify important dierences
between corrosion scales found in two dierent water distribution systems. Further studies are needed
to establish the role of corrosion scales in the mechanism of iron release from corroded pipes.
# 2001 Elsevier Science Ltd. All rights reserved
Key words}corrosion scales, corrosion, tubercles, iron oxides, enveloping layer

BACKGROUND

Iron and steel pipes have been used in water

distribution systems for over ve centuries. Since
the thermodynamically stable form of iron in contact
with atmospheric oxygen is ferric iron (Fe(III)),
exposure to aqueous conditions results in corrosion
of iron. Most water distribution pipelines have a
buildup of iron corrosion products inside the iron
pipes. Corrosion scales not only restrict the ow of
water (Larson and Sollo, 1967), but also degrade the
water quality. These scales are porous deposits that
are comprised of iron (hydr)oxide phases. Some ways
in which corrosion scales can adversely aect water
quality in drinking water distribution pipes are (a) as
a source of iron, which when released into water can
result in red water or colored water, (b) as a
source of high demand for chlorine and dissolved
oxygen (Tuovinen et al., 1984; Geldreich, 1990;
LeChevallier et al., 1993), (c) as an excellent breeding
ground for microbes, and (d) by adsorption and
accumulation of substances such as arsenic (Raven
et al., 1998), and radium (Field et al., 1995), which
can be released on modication of water quality.
Extensive research has been done to determine the
impact of water quality on the corrosion of new iron
pipes (Eliassen et al., 1956; Larson and Skold, 1958;

*Author to whom all correspondence should be addressed.

Tel.: +1-217-333-4700; fax: +1-217-333-9464; e-mail:
snoeyink@uiuc.edu

Stumm, 1960; Obrecht and Pourbaix, 1967; Pisigan

and Singley, 1987). However, the interaction of
corroded iron (iron metal surface having corrosion
scales) with water has not been studied in depth. The
response of corroded iron to changes in water quality
is expected to be a strong function of the physicochemical characteristics of the corrosion scales, and is
dierent from classical electrochemical corrosion of
iron.
Several water quality parameters can inuence the
formation of corrosion scales in iron/steel pipes in
water distribution systems. These include pH, alkalinity, buer intensity, dissolved oxygen (DO), water
ow characteristics, temperature, water treatment
practices, application of an inhibitor, and uctuations in water quality. Cornell and Schwertmann
(1996) have categorized the formation mechanisms of
iron oxides and oxyhydroxides into (a) direct precipitation from solution, and (b) topotactic phase
transformations from one solid phase into another by
internal rearrangements of atoms.
Scale structure and composition
Corrosion scales originate at the pipe wall and
grow radially inwards towards the pipe center. The
pipe metal is the original source of iron and the scales
are porous deposits with a distinct, layered structure.
Often a dense, shell-like layer has been observed near
the surface of the scale that is in contact with water.
Magnetite and goethite were reported as the

2961

2962

P. Sarin et al.

iron-containing phases comprising this dense layer

(Kolle and Rosch, 1980; Sontheimer et al., 1981).
Compounds usually found in iron corrosion scales
include
goethite
(a-FeOOH),
lepidocrocite
(g-FeOOH), magnetite (Fe3O4), siderite (FeCO3),
ferrous hydroxide (Fe(OH)2), ferric hydroxide
(Fe(OH)3), ferrihydrite (5Fe2O3
9H2O), green rusts
III
(e.g., FeII
4 Fe2 (OH)12(CO3)) and calcium carbonate
(CaCO3) (Benjamin et al., 1996). Depending on the
composition of the water next to the pipe surface,
corrosion of iron may result in the formation of
relatively soluble Fe(II) solids such as siderite and
ferrous hydroxide. Inside the scale, close to the pipe
wall, iron phases in the oxidation state Fe(II) are
expected. Ferrous hydroxide is formed under reducing conditions. However, in the presence of carbonic
species, siderite (FeCO3) is the stable ferrous solid
(Singer and Stumm, 1970). Studies by Baylis (1926),
Benjamin et al. (1996) and Sontheimer et al. (1981)
have documented the presence of siderite as a
corrosion product.
Green rusts are hydrated ferrousferric com2
pounds having Cl or SO2
4 or CO3 anions (Olowe
et al., 1988; Drissi et al., 1994; Genin et al., 1997;
Refait et al., 1998). These compounds are formed
during the partial oxidation of ferrous hydroxide in
the presence of other anions (Refait et al., 1997).
Touvinen et al. (1980) found green rust as the
primary crystalline form of iron in cast iron
corrosion scales in the Columbus, Ohio, distribution
system. The degree of iron oxidation in green rust is
variable and an Fe(II)/Fe(III) ratio of 4.00.8 was
observed.

Under anoxic conditions, the amount of iron

released could depend upon the solid phase in the
scale that controls solubility and solution composition such as pH and the total carbonate concentration. For example, when siderite is present and the
total carbonate concentration is 103 M, it will
control Fe(II) solubility below pH 10.5, and ferrous
hydroxide will control solubility above pH 10.5 at
258C (Snoeyink and Jenkins, 1980). Results from
Genin et al. (1998) showed that green rust controlled
iron solubility in soil solutions and this phase may
also be important in corrosion scales. A conceptual
model of protective scale formation, proposed by
Kolle and Sontheimer (1977) and further explained
by Sontheimer et al. (1981), highlighted the critical
reactions that might occur within the scales. Later,
Kuch (1988) proposed a mechanism of non-steadystate corrosion of iron by reduction of ferric oxide
scales with subsequent release of Fe(II) ions to the
solution.
The primary objective of this study was to
characterize the physical properties and chemical
composition of corrosion scales, and to identify the
dierences between scales from dissimilar water
distribution systems. The subjects of this study were
iron/steel pipes that had been in service in drinking
water distribution systems for several decades.
The information obtained should be useful for
understanding reasons for the success and failure
of colored water and scale formation control
procedures.

MATERIALS AND METHODS

Corrosion and iron release

Corrosion is the oxidation of metallic iron, and
this process releases iron into solution, or results in
iron scales. Corrosion may uniformly attack a metal
surface (uniform corrosion) or it may be focused at
specic sites. While uniform corrosion results in the
development of uniform scales, localized corrosion
can produce growing mounds of corrosion products
(Obrecht and Pourbaix, 1967), called tubercles
(Larson and King, 1954; Larson and Skold, 1957).
Iron release refers to the transport of iron from the
corrosion scale or the metal surface to the bulk water,
either in soluble form or as particulate matter. Iron
release from corroded iron surfaces can be brought
about by (a) corrosion of metal (b) hydraulic
scouring action of owing water and (c) dissolution
of corrosion scale components. In water distribution
systems iron release is a cumulative eect of all the
above. Whereas corrosion rate is usually measured
by weight loss from the metallic iron, iron release is
measured as the concentration of iron in the bulk
water. The rate of iron release often bears no simple
relationship to the rate of iron corrosion (Benjamin
et al., 1996).

Corrosion scales from two dierent water distribution

systems were analyzed. A 70-year old galvanized steel pipe,
1.5-m long, and 3.2-cm diameter, was excavated from the
Northern Illinois Water Corporations (NIWC) water
distribution system in Champaign, IL, USA. The pipe had
been in the distribution system since the early 1920s. The
inner surface of the pipe had a semi-uniform corrosion layer
with a thickness of 25 mm. For comparison, corrosion
scales from 90 year old, 20.3-cm diameter cast iron pipes
from the Metropolitan Water Resources Authority
(MWRA), Boston, MA were also analyzed. The cast iron
pipes from MWRA were heavily tuberculated, occluding up
to 76% of the cross-section originally available for ow.
Table 1 compares various elements of water quality of the
MWRA and NIWC distribution systems and the pipes that
were studied from these systems. The quality of water in the
two distribution systems was dierent. Although the water
chemistry data for these systems is incomplete for the
duration that the pipes were in use, data from recent years
(19961998) show that MWRA water had an alkalinity of
1035 mg/L as CaCO3, and a pH of 79. The NIWC water,
in comparison, had signicantly higher alkalinity (135
150 mg/L as CaCO3) and pH close to 9.0. The Ca and Si
content of NIWC water were signicantly higher than
MWRA water.
Experimental pipe loops were set up using six sections of
a cast iron pipe from MWRA, and a galvanized pipe from
NIWC. The inuent water was carefully controlled, and the
euent water quality was monitored for both short and
long-term iron release studies. Corrosion scales were
sampled for analysis from these pipes (see Table 2). The

Corrosion scales in old iron pipes

2963

Table 1. Comparison of MWRA and NIWC distribution systems

MWRA distribution system

NIWC distribution system

Water quality
pH
DO
Total alkalinity
Ca
Si
Cl
SO2
4 (mg/L)

78
78.5 mg/L
1035 (mg/L as CaCO3)
3.6 mg/L
2.22.3 mg/L as SiO2
12.4 mg/L
5.6 mg/L

8.9
14 mg/L
135150 (mg/L as CaCO3)
1116.5 mg/L
78 mg/L as SiO2
14 mg/L
12 mg/L

Pipe
Material
Diameter
Age in service

Cast iron
20.32 cm (8 inch)
90 years (approx)

Galvanized steel
3.18 cm (1.25 inch)
70 years (approx)

Table 2. Details of corrosion scale samples

Sl. No.

Sample code

Description of sample

1
2
3
4
5
6
7
8

NIWC-A
NIWC-B
Boston#1
Boston#2
Boston#3
Boston#4
Boston#5
Boston#6

As received sample of galvanized pipe from NIWC

Sample from the NIWC pipe after 1 year in pipe loop
Cast iron pipe scale sample from Boston pipe # 1
Cast iron pipe scale sample from Boston pipe # 2
Cast iron pipe scale sample from Boston pipe # 3
Cast iron pipe scale sample from Boston pipe # 4
Cast iron pipe scale sample from Boston pipe # 5
Cast iron pipe scale sample from Boston pipe # 6

NIWC-A scale sample was removed from the as-received

pipe, whereas the NIWC-B scale was taken from the same
pipe after conducting short-term iron release studies for
about 1 year in the pipe loop. The inuence of stagnation
time, dissolved oxygen concentration, and pH on iron
release was observed before the NIWC-B sample was
removed.
Scale analyses
The scales were dried under vacuum at about 208C and
were characterized by several experimental techniques. In
order to conrm if drying the scales modied any chemical
species present in the wet scales, the concentrations of Fe(II)
and Fe(III) in the scales were compared before and after
drying.
Dry corrosion scales: Pipe sections with corrosion scales
were rst dried under vacuum and room temperature
conditions. The dried scales were then removed from the
pipe wall and dierent types of samples were prepared.
Cross-section samples: The dried scales were impregnated
with an epoxy resin (Epo-Thin, Buehler Ltd., Lake Blu,
IL) under vacuum, and sectioned by using a slow-action
diamond blade. Epoxy impregnation was often repeated for
these porous scales. The cross-sectioned samples were
polished to a 1 mm surface nish on Buehlers ECOMET
III polisher/grinder (Buehler Ltd., Evanston, IL) using a
diamond paste (Buehler METADI Aerosol Spray Diamond
Compound 1 mm; Buehler Ltd., Lake Blu, IL). Petrographic slides were also prepared for few specimens
(Wagner Petrographic, Provo, UT).
Surface samples: A small part of the top surface of the
dried tubercle was studied using scanning electron microscopy (SEM). This surface layer, referred to as the top
surface of the scale, is present on top of the dense shell-like
layer, and comes in direct contact with the owing water.
Powder samples: Samples of the entire tubercle were
powdered with an alumina mortar and pestle until they
passed through a 150 mm a mesh sieve. It was possible to

remove the distinct shell-like layer from the MWRA scales

using a razor blade, and separate powder samples were
prepared for this layer. In spite of all the precautions taken
during sampling, it was not possible to sample this dense
layer exclusively. Therefore, these powder samples include
both the dense shell-like layer and the top surface layer
and a sux -TSS (top surface and shell) is used to identify
these samples.
Characterization of corrosion scale microstructure: Several
techniques were employed to study the physico-chemical
characteristics of the corrosion scales.
(a) Physical structure: A table-top optical microscope
(Nikon SMZ-2 T, Nikon, Inc., Japan) was used to study the
peterographic slides and other cross-section samples. This
analysis provides a macro-view of the corrosion-scale
structure.
(b) Bulk density: The dried scale samples were rst
weighed and then their volume was determined. In order to
determine bulk density (mass/volume, including pores), the
scale sample was covered with a very thin coat of paran
wax to prevent water from penetrating the scales. Scale
volume was calculated from the amount of water displaced
when immersed in water (Archimedes method). Mettlers
AE166 electronic weighing balance (Mettler Instrument
Corp., Hightstown, NJ, USA), along with accessories for
density measurement, was used for these experiments.
(c) True density: True density, mass/volume of solid
excluding pores, of powdered samples of corrosion scales
was determined using the AccuPyc 1330 Pycnometer
(Micromeritics, Norcross, GA) to measure the volume of
helium gas displaced by a known mass of powder sample.
(d) Surface area and pore size: The ASAP 2400 analyzer
(Micromeritics, Norcross, GA) was used to determine the
specic surface area of powder samples by N2 gas
adsorption and desorption. The samples were degassed at
1208C prior to analysis.
(e) Elemental composition: Powder, top surface and
polished cross-section samples were analyzed using SEM
and energy dispersive spectroscopy (EDS) techniques. A
LaB6 lament, multipurpose, analytical scanning electron

2964

P. Sarin et al.

microscope (Zeiss DSM 960, Gottingen, Germany) and

SEM Hitachi S-530 ( Hitachi, Conroe, TX)) equipped with a
tungsten (W) lament were used for the SEM/EDS studies.
The EDS results were limited to elements with atomic
number 11 or greater.
(f ) Crystalline phase composition: A RIGAKU (D-MAX
II) X-ray powder diractometer (Rigaku/USA Inc., Danvers, MA) with a Cu Ka source (l 0:1540598 nm) and a
single crystal monochromator in the diracted beam path,
was used for acquiring powder diraction data. Multiple
scans were performed to improve resolution. The X-ray
diraction (XRD) pattern was analyzed by the Rietveld
method (Rietveld, 1969) using the General Structure
Analysis System (GSAS) software from Los Alamos
National Laboratory (Von Dreele and Larson, 1997).
Crystalline-phase composition of the corrosion scales was
quantitatively determined following the procedure of Hill
and Howard (1987).
(g) Fe(II)/Fe(III) content: The Fe(II)/Fe(III) ratio of the
powdered corrosion scale was measured by wet chemistry
methods. Samples were dissolved in an acidic solution
(20 mL of concentrated HCl (Fisher Scientic Certied
ACS) in 300 mL distilled and de-ionized (DDI) water.
The ferrous and total iron concentrations were determined
using
bathophenanthroline,
following
the
method proposed by Lee and Stumm (1960). A Beckman
DU 7500 Spectrophotometer (Beckman Coulter Inc., Fullerton, CA) was used to analyze the absorbance of the
prepared samples.
Wet corrosion scales: Scales from three cast iron pipes
from MWRA being used in pipe loop studies were sampled
for these analyses. The water ow through these pipes was
stopped and small parts of the scales were scraped from top,
middle and bottom of the tubercles close to the end of the
pipes. Samples were immediately dissolved in an acidic
solution (20 mL of concentrated HCl in 300 mL DDI water),
avoiding exposure to atmospheric environment, and their
Fe(II)/Fe(III) content was analyzed. Scales from the NIWC
pipe section were sampled in a similar way at two locations.
Since the scales in NIWC pipes were much thinner in
comparison to MWRA scales, they were analyzed as an
entire scale only.

Fig. 1. Optical microscopic image of the cross-section of the

corrosion scales from NIWC showing a dense exterior layer
and a porous interior.

Fig. 2. Optical microscopic image of the cross-section slide

of the corrosion scales from NIWC showing dense shelllike layer, outermost layer at the waterscale interface, and
the porous interior usually present in corrosion scales found
in iron/steel drinking water distribution pipes.

RESULTS AND DISCUSSION

The physico-chemical characteristics of the corrosion scales can determine the ensuing chemical
reactions when corroded pipe surfaces are exposed to
aqueous conditions. The important features of the
structure and composition of MWRA and NIWC
scales are presented in the following sections.
Physical characteristics of corrosion scales
Corrosion scales provide resistance to further
corrosion of the underlying metal, and porosity is
one factor dictating their eectiveness. Porosity can
determine the ease with which ions migrate within the
scale layers (Benjamin et al., 1996).
The cross-section samples of MWRA and NIWC
scales were viewed under an optical microscope to
study the physical structure of the scales. A shell-like
layer was observed near the top of the tubercle, i.e.,
the section of the tubercle in close contact with water.
The part of the corrosion scale near the pipe wall
was very porous, and had voids as large as 12 mm,

observable by eye. An optical micrograph of NIWC

corrosion scale is shown in Fig. 1. The dense shelllike layer isolated the porous interior of the scales
from bulk water. The interface between the corrosion
scale and water was not sharp (see Fig. 2). A
combination of Archimedes method and true density
measurements of powders (pycnometry) was used
to evaluate porosity of scales (see Table 3). The
measured bulk density was 2.22.3 and 1.72.2 g/cm3
for NIWC and MWRA scales, respectively. The
calculated porosity values were 4043 and 4454%
for NIWC scales and MWRA scales, respectively.
Corrosion scales are heterogeneous and variation in
tubercle structure between two sampling sites can
explain the observed range in bulk density and
porosity for MWRA scales. The lower density and
higher porosity of the MWRA scales in comparison
to NIWC scales may have been caused by dierences
in water quality, dierent water ow rates, and
microorganism growth.

Corrosion scales in old iron pipes

2965

Table 3. Density and porosity determination of the corrosion scales

Sample ID

True density (g/cm3)

Bulk density in (g/cm3)

Porosity (%)

NIWC-A
NIWC-B
Boston#1
Boston#2
Boston#3
Boston#4
Boston#5
Boston#6

3.89
3.93
3.72
3.92
3.78
3.60
3.59
3.67

2.20
2.34
1.75
1.80
1.73
2.01
1.99
2.22

43.30
40.51
53.02
54.18
54.33
44.13
44.43
39.56

Table 4. Surface area and porosity of powders from dry scale samples
Sample ID

NIWC-A
NIWC-B
Boston#1-TSSb
Boston#1
Boston#2-TSS
Boston#2
Boston#3-TSS
Boston#3
Boston#4-TSS
Boston#4
Boston#5-TSS
Boston#5
Boston#6-TSS
Boston#6
a
b

Pore size in A

Volume in cm3/g (  103)

Surface area (sq. m/g)

BET

SA of pores
1730008A
dia (BJHa)

m-pore
area

Vol. of pores
1730008A dia (BJHa)

m-pore
vol

Avg pore dia

(BET)

Avg pore dia

(BJHa)

154
89
60
43
50
32
95
43
79
82
49
53
41
54

85
46
54
47
46
36
73
45
64
70
41
57
33
49

28
32
8
1
8
0
4
1
12
9
7
0
3
1

90
70
140
160
110
120
160
180
190
240
140
240
130
170

12
14
4
0
4
0
1
1
5
4
3
0
1
0

32
39
93
130
91
142
63
125
89
97
98
160
114
117

42
64
106
138
99
140
87
163
122
139
139
165
132
140

Calculated using desorption data by Brunauer, Jura and Harkins (BJH) method.
Top layers of the Boston samples were analyzed separately also, and they have a sux of -TSS (top surface and shell) at the end of their
code.

The physical characteristics of powders prepared

by grinding dry scales are important for comparing
scales from dierent systems. Results of the nitrogen
adsorption/desorption studies (see Table 4) show that
(a) NIWC scale powders had very high micropore
surface area (2832 m2/g) in comparison with powders from MWRA scales (negligible to 1.5 m2/g),
(b) NIWC scale powders had a higher Brunauer,
Emmett and Teller (BET) specic surface area (89
154 m2/g) than did the MWRA samples (3254 m2/g),
and (c) the average pore diameter (by BET) of NIWC
scale powders (3239 A) was less than one-third of
MWRA scale powders (129160 A). The Boston #4
sample showed the highest micropore area, highest
surface area, and smallest average pore size of all the
MWRA samples. This was probably due to an
experimental error. For MWRA samples, powders
from the top surface and shell (shaded rows and
Sample ID with sux -TSS) samples had higher
surface areas, higher micropore areas, and smaller
average pore diameters when compared to powders
prepared from the entire corrosion scale from the
same MWRA sample.
Chemical composition of corrosion scales
The chemical composition of corrosion scales was
characterized using EDS, XRD and wet-chemistry

methods. The elemental composition of the scales as

identied by EDS measurements (see Table 5)
showed that they were primarily comprised of iron
(Fe). The presence of Zn in the NIWC scales was
possibly from the galvanization layer on the steel
pipe. However, copper (Cu) was not present in the
pipe metal or the galvanization layer, and the water
was the only possible source. Therefore, the appearance of Cu in iron corrosion scales may only be
explained by absorption from water. The high
concentration of silicon (Si) in the top surface layer
of the NIWC scales is most likely due to its high
concentration in the NIWC water supply (Table 1).
Although elemental analysis provides useful information about the distribution of elements in
corrosion scales, it is the compounds which they
form that determine their reactivity. Since iron oxides
readily form stable crystalline phases, X-ray diraction was the preferred method for compound
identication. The XRD patterns were acquired for
powder samples and the crystalline phases present in
scales were identied. Additionally, quantitative
phase composition information was also extracted
from these patterns by using the Rietveld method
with GSAS software (Von Dreele and Larson, 1997).
In this method, the experimental X-ray diraction
pattern of the sample was least-squares-tted with a

2966

P. Sarin et al.
Table 5. EDS analysis of elemental composition of corrosion scale samplesa

Sample

Analysis of

Al

Si

Cl

Ca

Fe

Cu

Zn

NIWC-A

Surface
Bulk

}
1

17
8

1
}

}
1

}
}

5
1

65
7-

3
4

6
16

NIWC-B

Surface
Bulk

1
1

12
7

1
}

}
2

}
}

3
1

68
72

6
5

9
12

Boston#1

Surface
TSSb
Bulk

2
1
1

8
2
1

1
}
1

1
1
3

}
}
1

}
}
}

87
93
92

}
}
}

}
2
1

Boston#2

Surface
TSS
Bulk

1
1
4

4
2
1

1
}
}

1
}
2

}
}
}

1
1
}

89
92
9-

}
}
1

}
}
}

Boston#3

Surface
TSS
Bulk

3
1
1

6
2
1

1
}
1

1
1
2

}
}
}

}
1
}

88
93
93

}
}
}

1
1
1

Boston#4

Surface
TSS
Bulk

4
2
1

11
4
1

1
1
1

1
1
3

}
}
}

}
}
}

78
88
92

}
1
}

2
}
}

Boston#5

Surface
TSS
Bulk

2
1
1

6
2
1

}
}
}

1
2
3

5
3
3

1
}
}

78
91
91

}
}
}

2
}
}

Boston#6

Surface
TSS
Bulk

2
1
1

4
3
1

}
}
}

1
1
3

1
}
}

}
}
}

91
93
92

}
}
}

}
}
}

a
b

All values are in atom% with an error of  5%. The reported values are normalized and values less than 1 atom% are not reported.
Top layers of the Boston samples were analyzed separately and they have a sux of -TSS (top surface and shell) at the end of their code.

Table 6. Rietveld analysis of powder XRD data from MWRA and NIWC scales
Sample ID

w2

Magnetite (Wt.%)

Geothite (Wt.%)

Lepidocrocite (Wt.%)

NIWC-A
NIWC-B
BOS-1 TSS
BOS-1
BOS-2 TSS
BOS-2
BOS-3 TSS
BOS-3
BOS-4 TSS
BOS-4
BOS-5 TSS
BOS-5
BOS-6 TSS
BOS-6

2.76
2.79
2.41
2.19
2.38
2.67
2.39
2.22
2.53
2.41
2.33
1.99
3.61
2.41

24.3
20.7
37.2
34.0
46.2
35.5
40.5
35.7
31.0
23.3
44.4
30.5
31.6
27.6

75.7
79.3
43.9
59.4
33.6
60.6
43.4
59.8
49.9
71.5
41.9
59.0
58.5
67.3

0.0
0.0
18.9
6.5
20.2
3.9
16.0
4.4
19.1
5.2
13.7
10.5
9.9
5.1

Note: The samples from the top surface have -TSS as a sux in the Sample ID.

calculated diraction pattern. The diraction pattern

of standard crystal structures of compounds suspected of being present were input to the program,
and the amount of each phase was varied until the
calculated diraction matched the measured pattern
with a desired accuracy. The powder samples were
found to be crystalline and no amorphous phase was
detected. Since the corrosion scales were formed
under natural conditions, there was a strong possibility that the compounds present in corrosion scales
were impure. The use of diraction patterns for
standard crystal structures to calculate the simulated
diraction pattern was the probable reason for large
goodness of t parameter (w2 ) values obtained

(see Table 6). A sample Rietveld tting of an XRD

pattern is shown in Fig. 3. The plus (+) symbols
represent the experimental data and the grey line,
almost overlapping with the experimental data, is the
simulated diraction pattern. The three rows of
markers under the experimental and simulated
patterns show the peak positions for lepidocrocite,
goethite and magnetite, in that order from top row to
the bottom row. The curve at the bottom of the
graph is the dierence between the experimental and
the simulated patterns.
The quantitative phase composition of MWRA
and NIWC scales is summarized in Table 6. Our
results conrmed that goethite, magnetite and

Corrosion scales in old iron pipes

2967

Fig. 3. Rietveld tting of powder XRD studies on the corrosion scales from NIWC, Champaign, IL, for
quantitative analysis of phase composition. The three rows of markers correspond to the peak positions
for lepidocrocite (top row), goethite (middle row), and magnetite (bottom row), respectively.

lepidocrocite were the dominant constituents of iron

corrosion scales. This is in agreement with reports by
other researchers. Stumm (1960), Baylis (1926) and
Larson and King (1954) have also reported siderite as
a component mineral phase, but the scales from
MWRA and NIWC did not contain any siderite.
Moreover, lepidocrocite was conspicuously absent in
the NIWC scales. Higher concentrations of magnetite
in the top layers of MWRA scales is in accordance
with the nding of Benjamin et al. (1996) that the
percentage of magnetite increased with distance from
the pipe wall.
The shell-like layer was present in all of the
corrosion scales analyzed during this study. Its role
has been examined in considerable depth by others
(Kolle and Sontheimer, 1977; Kolle and Rosch, 1980;
Sontheimer et al., 1981). Goethite and magnetite
have been reported as the phases comprising this
dark, relatively hard and brittle layer (Benjamin
et al., 1996). Powder samples from top layers (surface
and shell) from MWRA scales were comprised of
magnetite, goethite and lepidocrocite. The top layers
come into contact with waters having high DO and
chlorine concentrations. The presence of lepidocrocite in top layers is consistent with this mineral phase
being formed by rapid oxidation of ferrous hydroxides (Cornell and Schwertmann, 1996).
Synthesis studies have shown that goethite formation predominated over lepidocrocite formation at a
slow oxidation rate (Benjamin et al., 1996). A high
pH and a slower rate of oxidation favored formation
of magnetite over lepidocrocite (Cornell and
Schwertmann, 1996). Since the MWRA and NIWC

pipes have been in service for several decades and

have presumably witnessed a wide variation in water
quality during use, it is not possible to give the
conditions that were responsible for the formation of
lepidocrocite in MWRA pipes and its absence in
NIWC system pipes.
Dry vs. wet scales
The analyses reported above were for dried
samples. In order to determine if sample preparation
(vacuum drying at room temperature) had modied
the scale composition, small amounts of wet scale
samples were dissolved in 1 M HCl, and the solution
was analyzed for Fe(II) and Fe(III) concentrations by
the bathophenanthroline method. Results showed
that wet scales had a relatively higher Fe(II) content
than the dry scales (Tables 7 and 8). Analyses of three
MWRA samples showed a decrease in Fe(II) content
towards the top of the tubercles, reconrming that
the oxidation state of iron increased with distance
from the pipe wall (Table 7). These samples also
showed a high variability from sample to sample,
indicating non-uniformity of scale composition.
Powder samples used for XRD studies were also
analyzed for their Fe(II)/Fe(III) content to check the
correctness of the percentage phase composition as
obtained by the Rietveld method. While goethite and
lepidocrocite have iron in the Fe(III) state only,
magnetite (Fe3O4) has one Fe(II) and two Fe(III) in
each molecule. Calculation of Fe(II) content from the
phase composition determined from XRD results
correlated well with the results of the wet chemical

2968

P. Sarin et al.
Table 7. Fe(II)/Fe(III) analyses of wet corrosion scales
Sample details

Ferrous (%)

Ferric (%)

Sample ID

Sampling position

NIWC-A

Entire scale

30.7
29.2

69.3
70.8

Boston#1

Top
Middle
Bottom

25.9
22.3
52.6

74.1
77.7
47.4

Boston#3

Top
Middle
Bottom

3.5
94.5
92.2

96.5
5.5
7.8

Boston#6

Top
Middle
Bottom

28.3
79.9
80.6

71.8
20.1
19.4

Table 8. Comparison of calculated Fe(II) and Fe(III) content from

XRD studies with wet chemistry results for dry powder samples
Fe2+ and Fe3+ (wt.%)

Sample ID

Calculated from XRD

NIWC-A
NIWC-B
BOS-1 TS
BOS-1
BOS-2 TS
BOS-2
BOS-3 TS
BOS-3
BOS-4 TS
BOS-4
BOS-5 TS
BOS-5
BOS-6 TS
BOS-6
a

Wet chemistrya

Fe(II)

Fe(III)

Fe(II)

Fe(III)

9.0
7.7
13.5
12.4
16.6
12.9
14.6
13.0
11.3
8.6
16.0
11.2
11.6
10.2

91.0
92.3
86.5
87.6
83.4
87.1
85.4
87.0
88.7
91.4
84.1
88.8
88.5
89.9

14.2
15.5
15.6
11.6
13.9
13.6
13.1
12.9
16.9
12.6
14.0
10.3
11.8
9.3

85.8
84.5
84.4
88.4
86.1
86.4
86.9
87.1
83.1
87.4
86.0
89.7
88.2
90.7

Values averaged for three measurements

analysis on the same powders (Table 8). This veried

the accuracy of the Rietveld analysis of XRD
patterns.
The observance of higher Fe(II) concentrations in
wet scales compared to dry scales indicated that
Fe(II) phases were converted to Fe(III) phases during
drying. The dry samples had ferrous ions present
only in the form of magnetite (Fe3O4), and the
conversion of a phase such as Fe(OH)2, to magnetite
by oxygen according to the following reaction, could
account for the observed change:
1
2 O2

3FeOH2 ! Fe3 O4 3H2 O;

CONCLUSIONS

The complex microstructure of the corrosion scales

from old iron/steel water distribution pipelines was
analyzed using several analytical techniques. Corrosion scales were identied as porous deposits of iron
oxide or oxyhydroxide phases. A distinct shell-like
dense layer was present near the scalewater interface
in all the scales analyzed. An apparent pattern was
observed in the distribution of phases within the
scales. The top layers of the corrosion scales had
higher concentrations of magnetite and lepidocrocite
phases. Fe(II)/Fe(III) analyses of wet and dry scales
from the two systems illustrated that wet scales had
signicantly higher amounts of Fe(II). Important
dierences were established between the corrosion
scales from MWRA and NIWC. The high concentration of Si in the top layer of NIWC scales was
probably due to the higher concentration of this
element in the NIWC water. Magnetite, goethite and
lepidocrocite were the primary crystalline phases
present in MWRA corrosion scales, while only
magnetite and goethite were found in NIWC scales.
MWRA scales were more porous than NIWC scales.
Comparison of the physical characteristics of the
powdered scales showed that particles from MWRA
scales had lower surface areas and insignicant
microporosity in comparison to the NIWC scale
particles. While we do not understand the exact
reasons eecting such particle formations in the two
systems, we believe that the dierences are noteworthy.
Our studies have been able to identify measurable
physico-chemical characteristics of corrosion scales.
These characteristics are thought to be important
factors that aect the release of iron and the
degradation of water quality in drinking water
distribution systems. However, research is needed
to show the mechanism of this release and to develop
ecient water quality control strategies.
Acknowledgements}The authors acknowledge the use of
facilities of the Center of Microanalysis of Materials at the
Materials Research Laboratory, UIUC and the NSF Center
for Advanced Cement Based Materials at UIUC. Prof. G. P.
Wirtz and Prof. H. H. Chen at Department of Materials
Science and Engineering at UIUC are thanked for some
insightful discussions, as is C. J. Johnson (UIUC) for
helping with the wet chemistry analyses, and Mr. M Hayes
(Black and Veatch, Inc.) for operating the MWRA pipe
loops. This research was supported by the American Water
Works Association Research Foundation (AWWARF) and
the Metropolitan Water Resources Authority (MWRA),
Boston, MA.

DG8 272 kJ=mol

This explanation is reasonable because the oxidation
reaction is thermodynamically favored at ambient
conditions and Fe(OH)2 may exist between the shelllike layer of the tubercle and the pipe wall. Other
similar reactions involving green rust cannot be ruled
out, however.

REFERENCES

Baylis J. R. (1926) Prevention of corrosion and red water.

J. Am. Water Works Assoc. 15(6), 598631.
Benjamin M. M., Sontheimer H. and Leroy P. (1996)
Corrosion of iron and steel. In Internal Corrosion of
Water Distribution Systems. Cooperative Research Report, AWWA Research Foundation, Denver, CO.

Corrosion scales in old iron pipes

Cornell R. M. and Schwertmann U. (1996) The Iron Oxides,
573pp, VCH Publishers, Weinhein.
Drissi H., Refait P. and Genin J. M. R. (1994) The
oxidation of Fe(OH)2 in the presence of carbonate ions:
structure of carbonate green rust one. Hyperne Interactions 90, 14.
Eliassen R., Pereda C., Romeo A. J. and Skrinde R. T.
(1956) Eects of pH and velocity on corrosion of steel
water pipes. J. Am. Water Works Assoc. 48(8), 1005
1018.
Field R. W., Fisher E. L., Valentine R. L. and Kross B. C.
(1995) Radium-bearing pipe scale deposits}implications
for national waterborne radon sampling methods. Am. J.
Public Health 85(4), 567570.
Geldreich E. E. (1990) Internal corrosion and deposition
control. In Water Quality and Treatment. McGraw Hill,
New York.
Genin J. M., Refait P., Olowe A. A., Abdelmoula M., Fall I.
and Drissi S. H. (1997) Identication of green rust
compounds in the aqueous corrosion processes of steels;
the case of microbially induced corrosion and use of 78 K
CEMS. Baltzer. Hyperne Interactions 112, 14.
Genin J. -M. R., Bourrie G., Abdelmoula M., Jarezic A.,
Refait P., Maitre V., Himbert B. and Herbillon A. (1998)
Thermodynamic equilibria in aqueous suspensions of
synthetic and natural Fe(II)Fe(III) green rusts: occurences of the mineral in hydromorphic soils. Environ. Sci.
Technol. 32(8), 10581068.
Hill R. J. and Howard C. J. (1987) Quantitative phase
analysis from neutron powder diraction data using the
Rietveld method. J. Appl. Crystallogr. 20, 467474.
Kolle W. and Rosch H. (1980) Untersuchungen an
RohrnetzInkrustierungen unter mineralogischen Gesichtspunkten. Vom Wasser 55, 159178.
Kolle W. and Sontheimer H. (1977) Untersuchugen zur
Schutzschichtbildung in Gussrohren. Vom Wasser 49,
277294.
Kuch A. (1988) Investigations of the reduction and reoxidation kinetics of iron(III) oxide scales formed in
waters. Corros. Sci. 28(3), 221231.
Larson T. E. and King R. M. (1954) Corrosion by water at
low ow velocity. J. Am. Water Works Assoc. 46, 19.
Larson, T. E. and Skold, R. V. (1957) Corrosion and
tuberculation of cast iron. J. Am. Water Works Assoc. 49,
12941301.
Larson T. E. and Skold R. V. (1958) Laboratory studies
relating mineral quality of water to corrosion of steel and
cast iron. Corrosion 14(8), 285t288t.
Larson T. E. and Sollo Jr. F. R. (1967) Loss in water main
carrying capacity. J. Am. Water Works Assoc. 59(12),
15651572.
LeChevallier M. W., Lowry C. D., Lee R. G. and Gibbon
D. L. (1993) Examining the relationship between iron

2969

corrosion and disinfection of biolm bacteria. J. Am.

Water Works Assoc. 85(7), 111123.
Lee G. F. and Stumm W. (1960) Determination of ferrous
iron in the presence of ferric iron with Bathophenanthroline. J. Am. Water Works Assoc. 52, 15671573.
Obrecht M. F. and Pourbaix M. (1967) Corrosion of metals
in potable water systems. J. Am. Water Works Assoc. 59,
977992.
Olowe A. A., Genin J. M. R. and Bauer P. (1988) Hyperne
interactions and structures of ferrous hydroxide and
green rust II in sulfated aqueous media. Hyperne
Interactions 41, 501504.
Pisigan Jr. R. A. and Singley J. R. (1987) Inuence of buer
capacity , chlorine residual and ow rate on corrosion of
mild steel and copper. J. Am. Water Works Assoc. 79(2),
6270.
Raven K. P., Jain A. and Loeppert R. H. (1998) Arsenite
and arsenate adsorption on ferrihydrite}kinetics, equilibrium, and adsorption envelopes. Environ. Sci.Technol.
32(3), 344349.
Refait P., Abdelmoula M. and Genin J. M. (1998)
Mechanisms of formation and structure of green rust
one in aqueous corrosion of iron in the presence of
chloride ions. Corro. Sci. 40(9), 15471560.
Refait P., Drissi S. H., Pytkiewicz J. and Genin J.-M. R.
(1997) The anionic species competition in iron aqueous
corrosion: role of various green rust compounds. Corro.
Sci. 39(9), 16991710.
Rietveld H. M. (1969) A prole renement method for
nuclear and magnetic structures. J. Appl. Crystallogr. 2,
6571.
Singer P. C. and Stumm W. (1970) The solubility of ferrous
iron in carbonate-bearing waters. J. Am. Water Works
Assoc. 68, 198202.
Snoeyink V. L. and Jenkins D. (1980) Water Chemistry.
Wiley, New York, NY.
Sontheimer H., Kolle W. and Snoeyink V. L. (1981) The
siderite model of the formation of corrosion-resistant
scales J. Am. Water Works Assoc. 73(11), 572579.
Stumm W. (1960) Investigation on the corrosive behaviors
of waters. J. Sanitary Eng. Division ASCE 86(11),
2745.
Tuovinen O. H., Button K. S., Vuorinen A., Carlson L.,
Mair D. and Yut L. A. (1980) Bacterial, chemical,
and mineralogical characteristics of tubercles in distribution pipelines J. Am. Water Works Assoc. 72(11),
626635.
Tuovinen O. H., Mair D. M. and Banovic J. (1984) Chlorine
demand and trihalomethane formation by tubercles from
cast iron water mains. Environ. Technol. Lett. 5, 97108.
Von Dreele R. B. and Larson A. C. (1997) GSAS}General
Structure Analysis System. The Regents of the University
of California, Los Alamos, NM.