You are on page 1of 30

Corrosion Science 48 (2006) 24802509

www.elsevier.com/locate/corsci

Atmospheric corrosion of copper and the colour,


structure and composition of natural
patinas on copper
K.P. FitzGerald, J. Nairn, G. Skennerton, A. Atrens *

Materials Engineering, School of Engineering, The University of Queensland, Brisbane QLD 4072, Australia

Received 16 November 2004; accepted 10 September 2005


Available online 21 November 2005

Abstract

This paper describes the results of atmospheric corrosion testing and of an examination of patina
samples from Brisbane, Denmark, Sweden, France, USA and Austria. The aim was threefold: (1) to
determine the structure of natural patinas and to relate their structure to their appearance in service
and to the atmospheric corrosion of copper; (2) to understand why a brown rust coloured layer
forms on the surface of some copper patinas; (3) to understand why some patinas are still black
in colour despite being of signicant age. During the atmospheric corrosion of copper, a two-layer
patina forms on the copper surface. Cuprite is the initial corrosion product and cuprite is always the
patina layer in contact with the copper. The growth laws describing patina formation indicate that
the decreasing corrosion rate with increasing exposure time is due to the protective nature of the
cuprite layer. The green patinas were typically characterised by an outer layer of brochantite, which
forms as individual crystals on the surface of the cuprite layer, probably by a precipitation reaction
from an aqueous surface layer on the cuprite layer. Natural patinas come in a variety of colours. The
colour is controlled by the amount of the patina and its chemical composition. Thin patinas contain-
ing predominantly cuprite were black. If the patina was suciently thick, and the [Fe]/[Cu] ratio was
low, then the patina was green, whereas if the [Fe]/[Cu] ratio was approximately 10 at%, then the
patina is rust brown in colour. The iron was in solid solution in the brochantite, which might be des-
ignated as a (copper/iron) hydroxysulphate. In the brown patinas examined, the iron was distributed
predominately in the outermost part of the patina.
 2005 Elsevier Ltd. All rights reserved.

*
Corresponding author. Tel.: +61 7 3365 3738; fax: +61 7 3365 3888.
E-mail address: a.atrens@minmet.uq.edu.au (A. Atrens).

0010-938X/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.corsci.2005.09.011
K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509 2481

Keywords: Atmospheric corrosion; Copper; Patina

1. Introduction

Patinas on copper have been studied for a long period of time extending over 200 years
[121,4956]. This study of copper patinas, and the complementary study of the atmo-
spheric corrosion of copper [2032], has provided a general understanding of the com-
pounds present in patinas and an understanding of the conditions leading to patina
formation. The present paper presents an analysis of the detailed structure and composi-
tion of a range of natural patinas from Brisbane (B) Australia and two separate ranges of
patinas from Europe. As became evident, the patinas are heterogeneous and metallurgi-
cally complex, and in all the cases examined t a two layer model of cuprite next to the
copper metal, and a green brochantite layer on top of the cuprite which is consistent with
the prior studies [921]. The aims of this present research work were as follows:

1. To determine the metallurgical structure of the natural patinas and to relate their struc-
ture to their appearance in service and to the atmospheric corrosion of copper.
2. To understand why a rust coloured layer forms on the surface of some copper patinas.
3. To understand why some patinas are still black in colour despite being of signicant
age.

2. Present knowledge

During natural weathering, copper goes through a number of stages. Salmon-pink is


the colour for clean copper with essentially no surface oxide as for example after acid
cleaning. After exposure to the atmosphere, copper rapidly turns to the more familiar
copper colour due to a thin surface oxide. On further exposure, the colour darkens
to brown and then to black as the oxide grows in thickness. These changes in colour
are all due to the oxide, cuprite which has the chemical formula Cu2O wherein the copper
is in the +1 oxidation state [9]. Subsequently, a green layer forms on top of the oxide layer
by further reaction with trace atmospheric impurities so that the copper is in the +2 oxi-
dation state. This green upper layer provides the typical patinated appearance [9]. Obser-
vations by the authors and discussions with interested eye witnesses have indicated that
some patinas aged from 50 to 150 years have rust-coloured streaks or have a rust coloured
outer layer; these patinas are typically in urban areas very close to dense trac. Examples
are evident on the copper domes at the Flinders Street Station in Melbourne, Australia
and on some copper Domes in the Banhofstrasse in Zurich Switzerland. A further example
used to be the dome of the Customs House in Brisbane until it was restored to its former
green colour by carefully abrading the outer several millimeters of the whole dome. These
rust-coloured features do not appear to be from rust but appear to be a natural reaction of
the patina to the local environment. Other old patinas, typically on old (typically 50250
years old) church spires, have a black appearance on vertical surfaces and surfaces that are
sheltered from washing by rain.
2482 K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509

On copper roofs, the green patina is usually thought to be aesthetically pleasing and is a
major reason why copper is used extensively throughout the world as a prestige roong
material. The time scale for patina formation varies with local environmental conditions,
and is reported to be between 20 and 70 years in Europe [11]. The wide scale degree of
atmospheric/industrial pollution (particularly of SO2) in the middle of last century in Eur-
ope correlated with a decrease in time to patination, whereas the more recent European
wide environmental policies to decrease atmospheric pollution levels including atmo-
spheric SO2 would be expected to increase the time to patination. In clean environments
that are dry for substantial periods of time, like Adelaide in South Australia or Canberra
in the ACT, it is not uncommon for copper roofs to be still at the black stage after 8090
years with very little of the green patina compound present.
Copper patinas are spatially heterogeneous containing voids and discontinuities. The
porosity can be readily demonstrated by the ability of patinas to absorb signicant
amounts of water [9,10]. Polished cross-sections of patinas from dierent locations and
ages were studied by Kalish [12]. The thickness of the outer layer varied from 10 to
84 lm, but on average did not exceed 4550 lm. The average thickness of the inner cuprite
layer was 17 lm.
X-ray diraction (XRD) [912,14,15,21] has revealed that the external green layer most
commonly consists of brochantite, CuSO4 3Cu(OH)2 and the inner layer of cuprite, Cu2O.
Brochantite was found on copper exposed to rural, urban, industrial and marine environ-
ments. Atacamite, Cu2Cl(OH)3, is also a major constituent of the outer layer in marine
exposures with high chloride levels. Graedel et al. [9] analysed the thermodynamic stability
of the various patina compounds in terms of Porbaix E-pH diagrams and solubilities and
showed that the copper hydroxysulphate phases are 100 times more stable (less soluble)
than the equivalent chloride and carbonate compounds, and 10 000 times more stable than
the equivalent nitrate compounds. Thus brochantite is nearly always the most common
compound in the external patina layer formed upon extended atmospheric exposure, even
in situations like the Statue of Liberty situated as it is in New York Harbour and sur-
rounded by salt sea water [9].
Patinas from the Statue of Liberty in the New York harbour and from the roof of the
AT&T Bell Laboratories, Murray Hill, NJ contained [9,10,15] (in addition to cuprite, bro-
chantite and atacamite) traces of antlerite, CuSO4 2Cu(OH)2, and posnjakite, CuSO4
3Cu(OH)2 2H2O. Posnjakite, a hydrated form of brochantite, was only found in the
younger patinas and was considered to be a transient phase. The presence of antlerite,
which is stable at a lower pH than brochantite, was attributed to the acidic precipitation
experienced in New York harbour.

3. Experimental method

Atmospheric corrosion testing followed ASTM standard G50-76 for Conducting


Atmospheric Testing of Metals. Test coupons were electrolytic tough pitch copper
(AS110A, 99.9% Cu) designated TCU, electrosheet (99.9% Cu) as produced by Copper
Reneries Pty Ltd. designated ESU and acid cleaned electrosheet designated EAU. Cou-
pons were 70 100 mm sheets. TCU coupons had a thickness 0.98 0.01 mm with an
average mass of 60.9 0.7 g. Electrosheet coupons had a of thickness 0.65 0.03 mm
with an average mass of 39.6 1.7 g. Each coupon was cleaned and degreased with a
hot alkaline degreaser to a no water break nish before being acid cleaned. The two
K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509 2483

stage acid cleaning treatment detailed in ASTM standard G1 required samples to be sub-
merged in a 10 vol% H2SO4 solution for 35 min at 25 C followed by submersion in a
solution of 120 ml H2SO4/30 g sodium dichromate/880 ml water for 510 s at 25 C. Trip-
licate samples were exposed for each time period (10 d, 3 m, 6 m, 1 y, 2 y, 3 y, 4 y). Cleaned
and weighed copper samples were mounted with porcelain insulators on galvanised expo-
sure racks at The University of Queensland, (UQ), St. Lucia, Brisbane (moderate urbane/
light industrial). Each sample was supported at only four points along each sample edge by
the porcelain insulators. Samples were inclined 30 from the horizontal and faced south.
After each exposure period, samples were removed and analysed for corrosion products
and corrosion rates. All samples were weighed after exposure to determine weight gain
due to adhering corrosion products. Two of the three triplicate samples were then acid
cleaned (using the sulfuric acid/dichromate solution) to remove the corrosion products
and re-weighed to determine weight loss and the corrosion rate. The corrosion products
on the third sample were characterised using optical metallography, XRD, SEM, EDS,
XPS and ICP.
Patina samples of ages ranging from 7 to 332 years, from dierent locations have been
obtained as summarised in Table 5. Patina samples from Brisbane, Queensland, Australia
had atmospheric exposures ranging from 7 to 142 years. Patinas from Denmark (D26 to
D332) had been exposed to the atmosphere for periods ranging from 26 to 332 years; the
initial year of exposure varied from about 1650 to 1956. A sample from the roof of AT&T
Bell Laboratories, Murry Hill, NJ, USA, designated AT44 had been exposed to the atmo-
sphere for 44 years. This sample was part of a previous study by Graedel et al. [9] and Fra-
ney and Thomas [10]. Samples were also available from France (FR30, SBR45, SSV45,
CC157), Stockholm Sweden (ARC117, CRS150, RS83), and a range of samples from
the Hofburg Imperial Palace in Vienna, Austria (HV65A to HV95D).
The surface appearance of the patina samples was examined using optical and SEM
metallography. Polished metallographic patina cross-sections were examined under
cross-polarised reected light. Samples were prepared by mounting cross-sections in Bueh-
ler Epo-thin epoxy resin and successive grinding from 240 to 600 grit silicon carbide paper
followed by 3 then 1 lm diamond paste using a cotton polishing lap. SEM micrographs
and EDS scans were obtained with a JEOL 6400 eld emission scanning electron micro-
scope operating at 10 kV; the samples were mounted on solid brass stubs using double
sided conducting carbon tape and were carbon sputter coated to reduce electron beam
charging eects. Linescans were obtained using a JEOL JXA-8800L Electron Probe Micro
Analyser (EMPA) operated at an accelerating voltage of 10 kV with a 70 nA probe cur-
rent. Line proles 200 lm in length were obtained using a dwell time of one second on
400 points, with a step size of 0.5 lm between points.
Characterisation of patina compounds was carried out using X-ray diraction (XRD).
Corrosion products were analysed in situ on the copper roong sample and also after they
had been scrapped from the sample. Samples were scanned between 10 and 65 at a scan
speed of 1/min using Cu Ka radiation from an X-ray source operating at 40 kV and
40 mA. The uPDSM search/match program using the PDF 1994 database was employed
to solve each pattern.
X-ray photon spectroscopy (XPS) was carried out using a PHI 560 spectrometer.
ICP (inductively coupled plasma atomic absorption spectroscopy) was used to measure
the elemental concentrations in the corrosion products on the atmospheric corrosion expo-
sure coupons and in the natural patinas by dissolving the corrosion products or the patina
2484 K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509

from the copper substrate and analysing the solute by ICP. A 50 50 mm sample was
immersed in 30 ml of 32% AR grade HCl in a 50 ml beaker for 90 min and this solution
was transferred to a 50 ml volumetric ask. A further 15 ml of HCl was used to wash the
copper substrate and the 50 ml beaker. This solution was transferred to the 50 ml volumet-
ric ask and the volume made up to the mark with HCl. Blank experiments were carried
out weighing SUQ45 as a function of time of exposure to HCl. These showed that more
than 99% of the patina weight dissolved within 10 min and after 20 min there was a linear
increase of weight loss with time corresponding to dissolution of copper from the metallic
substrate. An extrapolation of the weight loss data to zero time indicated that the extra
amount of copper dissolved from the substrate in 90 min of exposure to HCl corresponded
to 1.6% of the total copper in the SUQ45 patina.

4. Results

4.1. Atmospheric corrosion

In atmospheric corrosion testing, samples were exposed on open racks to the atmo-
sphere and then examined to determine the corrosion rate and the corrosion products.
In all cases there was a weight gain after exposure and moreover these was no apparent
macroscopic evidence of loss of corrosion products from the samples. All the exposed cou-
pons, even after 2 years of exposure, appeared copper coloured to dark brown; no trace of
green brochantite was visible to the naked eye. However, after the 10 day exposure, light
green crystal conglomerates were visible by optical microscopy. As the exposure time
increased, so too did the size and surface concentration of these green crystal conglomer-
ates. A photographic record was maintained for the skyward and ground ward surfaces
exposed for 10 d, 3 m, 6 m, 1 y and 2 y [7]. The size and density of the green crystal con-
glomerates was comparable on both the ground ward and skyward surfaces, with the sky-
ward surfaces for the 1 and 2 year exposures having slightly greater green conglomerate
coverage. For the 1 and 2 year exposure samples, there was a range of conglomerate sizes
ranging from submicron diameter up to 1030 lm, more or less evenly covering the whole
surface, with the coverage in the range 110%. The similarity of the skyward and ground
ward surfaces is attributed to the open nature of the exposure racks and the consequent
rain wetting of the ground ward surfaces of the coupons.
Table 1 presents the XRD identication of the corrosion products. Cuprite was present
on all samples and moreover the amount of cuprite increased with exposure time. XRD
indicated that brochantite was the stable external patina compound in agreement with
expectations from the literature [9]. Skyward and ground ward surfaces gave similar

Table 1
Corrosion products on TCU samples after atmospheric corrosion exposure at UQ, Brisbane (B)
Surface Corrosion products
3m 6m 1y 2y 3y 4y
Skyward C(M) C(M), B(t) C(M), P(m), B(m) C(M), B(m) C(M), B(m) C(M), B(m)
Groundward na na C(M), B(m) C(M), P(m), B(m) na na
C = cuprite Cu2O, B = brochantite, P = posnjakite, A = atacamite, M = major, m = minor, t = trace, na = not
analysed.
K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509 2485

XRD scans, with the skyward surfaces giving somewhat larger peaks for the basic copper
sulphates, brochantite and posjnakite. SEM and XPS examination of the sample surfaces
also revealed traces of chloride and moreover, the XPS analyses revealed comparable con-
centrations of S and Cl on both the skyward and ground ward surfaces. Examination of
both the skyward and ground ward sample surfaces using SEM metallography conrmed
that the brochantite occurred as discrete isolated clumps of crystals on the cuprite base
layer, with the brochantite islands increasing in number and size with increasing expo-
sure time. Fig. 1 is a SEM micrograph showing the skyward surface of a TCU sample after
two years exposure at UQ. SEM observations showed no signs of pitting underneath the
brochantite crystals suggesting that they precipitate from solution and did not grow into
the cuprite as a localised sulphate nest. Experiments carried out by FitzGerald et al. [6] on
the chemistry of brochantite precipitation support the model of brochantite precipitation
from the surface water lm.
These observations are interpreted as indicating that the atmospheric corrosion of cop-
per involves the formation of a layer of cuprite on the copper surface. Some of the cuprite
is further oxidised to Cu2+ and brochantite is precipitated on the cuprite surface.
Elemental concentrations of Cu, S and Fe per unit area of the corrosion products
retained on the combined skyward and ground ward sample surfaces of the TCU samples
exposed at Brisbane were determined by ICP, Table 2. The detailed corrosion product
analyses using optical and SEM metallography, XRD and XPS indicated that there was
slightly more corrosion product on the skyward surfaces for the 1 and 2 years exposure
samples. This dierence was suciently small to justify evaluation of the corrosion prod-
uct concentrations in Table 2 based on the premise that the corrosion product distribution
on the skyward surfaces was similar to that on the ground ward surfaces for each exposure
period. Knowledge of these elemental concentrations allows building up a detailed under-
standing of the corrosion reactions. Table 2 also includes the weight of the corrosion prod-
ucts, DWCP, determined from the weight of the samples measured after exposure and

Fig. 1. SEM micrograph of the skyward surface of TCU copper after 2 years exposure at UQ.
2486 K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509

Table 2
ICP analyses of 1 and 2 year atmospheric exposure samples at UQ, Brisbane
Sample and exposure Elemental composition (mg/cm2) Wtot DWCP [Fe]/[Cu]
(mg/cm2) (mg/cm2) (wt%)
Period, (y) [Cu] [S] [Fe]
11 0.878 0.0084 0.0011 1.032 1.018 1.7
12 0.938 0.0098 0.0007 1.108 1.082 0.9
13 0.970 0.0111 0.0018 1.149 1.095 2.0
21 1.328 0.0145 0.0009 1.570 1.600 0.8
22 1.378 0.0166 0.0015 1.637 1.592 1.1
23 1.453 0.0225 0.0020 1.752 1.627 1.1

before exposure. The total weight of the corrosion products was calculated, assuming that
cuprite and brochantite were the only corrosion products, using Eqs. (1)(3) as follows:
W bro SM bro =nS bro M S 1
W cup fCu  SnCu bro M Cu =nS bro M S gfM cup =nCu cup M Cu g 2
W tot W cup W bro 3
2
where Wbro is the weight of brochantite in the corrosion products (mg/cm ), Wcup is the
weight of cuprite in the corrosion products (mg/cm2), Wtot is the total weight of the cor-
rosion products (mg/cm2), [S] is the weight of sulphur in the corrosion products (mg/cm2),
[Cu] is the weight of copper in the corrosion products (mg/cm2), Mbro is the molecular
weight of brochantite, MS is the molecular weight of sulphur, MCu is the molecular weight
of copper, Mcup is the molecular weight of cuprite, nS_bro is the stoichiometric number of S
in brochantite, nCu_bro is the stoichiometric number of Cu in brochantite, and nCu_cup is the
stoichiometric number of Cu in cuprite.
There was good agreement between Wtot and DWCP.
The mass of copper in the runo (mg/cm2), WCu_runo, after 1 and 2 years of exposure
of TCU samples at Brisbane is given in Table 3, and was calculated using
W Cu runoff DW Cu  W Cu CP 4
where DWCu is the measured weight loss of metallic copper from the copper surface after
exposure and WCu_CP is the mass of copper in the corrosion products retailed on the sur-
face. These calculations indicated that after 1 year exposure, 82% of oxidised copper re-
mained on the sample surface with 18% lost to runo. After 2 years exposure, 91% of
oxidised copper remained on the sample surface with 9% lost to runo. These runo rates
are comparable with those from prior measurements: Kucera and Collins [33] in 12 month
exposures in Sweden measured runo rates of 48% at Vegagaten (precipitation pH 3.9),

Table 3
Copper mass balance for 1 and 2 year atmospheric exposure samples at UQ, Brisbane
Exposure period DWCu WCu_CP WCu_runo
(y) (mg/cm2) (mg/cm2) (mg/cm2)
1 1.14 0.93 0.21
2 1.52 1.38 0.14
Total copper weight loss, DWCu, equals weight of copper in corrosion products, WCu_CP, plus weight of copper in
runo, WCu_runo.
K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509 2487

29% at Mellbyleden (precipitation pH 4.0), and 17% at Floda (precipitation pH 4.3); He


[34] in 2.4 years exposures, measured runo rate of 20% for Stockholm, Sweden (precip-
itation pH 4.3). That is, our work showed that most of the corrosion products remained
on the sample surface consistent with the macroscopic observations of the atmospheric
samples and the locations where they were exposed to the atmosphere. Table 3 and the
prior works did however indicate that some copper was indeed lost in the runo. This
is consistent with the common observation of the blue staining observed at the base of
copper statues exposed to the atmosphere.
The average mass of copper per year in the runo, W Cu runoff , over the two year period
of exposure of the TCU samples at Brisbane was time averaged from the data in Table 3,
to be W Cu runoff 0:14  0:08 (mg/cm2/y). This is very similar to the yearly average runo
rate of 0.13 mg/cm2/y measured by He [34] for Stockholm Sweden. This similarity of the
present results to the literature provide condence in its validity.
The average quantity of cuprite oxidised, W Cu oxidised , over the two year period of expo-
sure of the TCU samples at Brisbane is given in Table 4 and was calculated from
W Cu oxidised W Cu bro W Cu runoff 5
where W Cu runoff is the average total mass of copper in the runo (from Table 3) and
W Cu bro is the average weight of copper in the brochantite given by
W Cu bro W Cu bro =time year and W Cu bro SnCu bro M Cu =nS bro M S 6
where the data for [S] comes from Table 2, and was also calculated from the examination
of Brisbane patina sections, Fig. 9. The oxidised cuprite as Cu2+ was either retained on the
surface in the 2+ valency state, W Cu bro , within brochantite or posnjakite, or washed from
the surface as runo, W Cu runoff . Table 4 indicates that the Cu2+ runo was 70% of the total
cuprite oxidised.
After removal of the corrosion products, there was a weight decrease indicative of the
loss of metallic copper from the samples. The weight loss was converted to a corrosion
rate, Fig. 2, which show the typical decrease in corrosion rate with exposure time consis-
tent with protective corrosion products. This is consistent with numerous atmospheric cor-
rosion tests which have shown a linear bi-logarithmic law between total corrosion
penetration, Pt, and time, t, as given by Eqs. (7) [35]
P t ktn 7a
log P t =t log k n  1 log t 7b
where Pt/t is the mean corrosion rate and k and n are constants which may be site specic.
Fitting Eq. (7b) to all the data of Fig. 2 yields values of n = 0.55 and k = 1.2 lm/y. These
values are comparable to values measured in prior longterm studies. Kucera et al. [26],

Table 4
Copper mass balance for 2 year atmospheric exposure samples at UQ, Brisbane
W Cu oxidised W Cu bro Eq. (6) W Cu bro Fig. 9 W Cu runoff

0.2 0.1 0.04 0.02 0.07 0.03 0.14 0.08


Average weight of cuprite oxidised (measured as mass of Cu/cm2 per year (mg/cm2/y)), W Cu oxidised , equals weight
of copper in brochantite, W Cu bro , plus weight of copper in runo, W Cu runoff . W Cu bro was evaluated from Eq. (6)
designated W Cu bro Eq. (6) and from the cross-section data of Fig. 9, designated W Cu bro Fig. 9.
2488 K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509

Fig. 2. Atmospheric corrosion rates for copper in Brisbane. Data up to 3 years is from atmospheric corrosion
testing, whereas corrosion rates at longer times were obtained by analysing roof samples, assuming a 30%
brochantite retention rate.

from 8 year exposure tests, evaluated the following n values: 0.75 (Prague), 0.66 (Usti),
0.78 (Hurnanovo), 0.66 (Kopisty) and 0.55 (Kasperske) all in Czechoslovakia; 0.86 (Stock-
holm), 0.70 (Ryda), 0.6 (Gotenburg) and 0.58 (Bohus) all in Sweden and 0.65 (Gallivare)
in Spain. Feliu and Morcillo [30], from 10 year tests in Spain, evaluated the following n
values: 0.33 and 0.52 (Alicante), 0.40 (Aviles), 0.48 (Barcelona), 0.62 (Sesatao), 0.41 (Ca-
diz), 0.37 (El Escrial), 0.48 (Madrid) and 0.61 (Zaragoza). Southwell and Bultman [25],
from 16 year tests in Panama, evaluated the following n values: 0.58 (Cristobal) and
0.42 (Mira Flores).

4.2. Natural patina samples

The patina samples had two very dierent surfaces. Typically the skyward surface was
coloured green more or less uniformly over the whole surface, although there were often
small black spots throughout the surface. Colour photographs are available in [7]. The
ground ward side was typically copper coloured, with the colour ranging from copper
brown to brown to black with increasing patina age.
Polished patina cross-sections were examined under cross-polarised reected light and
the skyward surfaces of samples were examined using SEM. In isolated areas of the older
Danish patinas (D324 and D332), atmospheric corrosion had penetrated along grain
boundaries to a depth of 0.080 mm which was about 1.6 times the average patina thick-
ness, Fig. 3. With this exception, in all cases the patina consisted of two layers as illus-
trated in Fig. 4 (a colour optical micrograph was included in [6]): cuprite next to the
copper metal, and a blue/green brochantite layer on top of the cuprite. This is consistent
with the prior analyses of natural patinas which have shown that there is an outer layer
consisting mainly of the basic copper sulphate, brochantite, and an inner layer of cuprite
[912,14,15,21]. Measurements were made from calibrated micrographs of the thicknesses
of the cuprite layer D0C , the thickness of the brochantite layer D0B , and of the total patina
K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509 2489

Fig. 3. Optical micrograph of D324 cross-section.

Fig. 4. Optical micrograph of CRS150 cross-section.

thickness D0T , and these results are presented in Table 5. Also given in Table 5 are the eval-
uations of cuprite layer thickness DIC and brochantite layer thickness DIB from the ICP
measurements as discussed below. There was excellent agreement between these indepen-
dent measurements of the layer thicknesses.
Fig. 5 presents thickness data for cuprite and brochantite from the controlled exposures
at Brisbane together with data from samples obtained from Brisbane copper roofs of var-
ious ages. A parabolic growth law can be seen for cuprite with the growth rate decreasing
from the initial rate of 20.2 lm/y at 15 year as the cuprite thickness increased steadily; the
cuprite thickness remained essentially constant at 6 lm for patina samples aged from 15 to
142 years. The brochantite thickness showed an approximately linear growth rate of
0.15 lm/y; the linear growth law is consistent with a non-protective layer and is consistent
with the open porous nature observed for the brochantite [9,10]. The decreasing growth
rate of cuprite between 1 and 15 years indicate an adherent dense oxide layer that is
responsible for the decreasing corrosion rate of copper exposed to the atmosphere. If there
was only the cuprite layer, its thickness would be expected to keep on increasing. The fact
that its thickness was constant between 15 and 142 years indicate that some cuprite was
actually lost and converted to brochantite. At 15 years the trend line indicates an expected
cuprite growth rate of 0.2 lm/y, which would correspond to a brochantite growth rate of
0.5 lm/y if all the copper ions were precipitated to form the brochantite layer. The actual
2490
K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509
Table 5
Sample designation, exposure age (y), sources of the patina, cuprite layer thickness D0C and DIC , brochantite layer thickness D0B and DIB , total patina thickness D0T ,
phases present, and patina colour
Sample Age Source of sample D0C D0B D0T DIC DIB Phases Col
(y) (lm) (lm) (lm) (lm) (lm)
C B P A O
HE7 7 Herston Med School, Brisbane (B), Qld 4.7 2 0.7 1 5.4 2 M t t Br
HE17 17 Herston Med School, Brisbane, Qld 8.1 4 4.1 5 12.2 5 M m t Br
HE27 27 Herston Med School, Brisbane, Qld 6.3 3 5.8 2 12.1 3 M m Gr
RUQ43 43 Richards Building, UQ, B 4.6 1 5.1 2 9.7 2 8.1 3 2.5 1 M m t Gr
AT44 44 AT&T Bell Laboratories, NJ USA 5.9 4 5.6 1 11.6 4 M M Gr
SUQ45 45 Steele Building, UQ, B 9.8 2 2.8 2 12.6 2 8.6 1 1.6 1 M m t Gr
HE54 54 Herston Med School, Brisbane, Qld 7.6 3 10.0 4 17.6 6 M m t Gr
CH142 142 Customs House, Brisbane, Qld 7.5 4 29.6 6 31.7 7 M M t Gr
D26 26 Broadcasting House, Copenhagen, DK 8.8 3 15.3 5 24.1 6 6.3 1 13.2 1 M M Gr
FR30 30 Saint-Die, France (F) 10.2 8 13.0 8 23.2 8 M M Gr
SBR45 45 Strasbourg, France (F) 5.0 24.5 29.5 M M Gr
SSV45 45 Saint Die (Vosges), France 9.5 22.6 32.1 M M Gr
D47 47 Vesterport, Copenhagen, DK 5.1 1 16.0 3 21.1 3 9.4 2 13.7 2 M M Gr
D55 55 Cathedral, Arhus, DK 13.6 1 13.8 1 Gr
D70 70 Fredriksberg Castle, Copenhagen, DK 7.2 3 25.3 4 32.5 5 7.4 1 19.7 1 t M Gr
D82 82 Town Hall Tower, Copenhagen, DK 8.2 2 25.3 3 33.5 4 13.9 2 18.9 1 M M Gr
D99 99 Marble Church, Copenhagen, DK 13.4 5 33.5 4 46.9 6 18.6 2 22.0 2 M M Gr
OSC100 100 Our Saviourss Church, DK 10.2 18.7 28.9 M M t Gr
AFC117 117 Adolf Fredricks Church, S 6.9 28.1 35.1 M M t Gr
CRS150 150 Riddarholmen Church, Stockholm, S 15.7 8 31.6 8 47.3 9 M M Gr
D153 153 Frederiksberg Castle, Copenhagen, DK 36.5 6 31.3 4 Gr
CC157 157 Cathedral de Chartres, F 9.6 22.2 31.8 M M t Gr
D230 230 Our Saviours, Copenhagen, DK 9.0 3 23.1 3 32.0 4 10.5 1 23.4 8 m M Gr
D324 324 Elsinore Castle, Elsinore DK 18.0 10 26.7 9 44.7 15 24.6 3 25.3 1 m M Gr
D332 332 Roskilde Cathedral, Roskilde DK 19.1 10 30.9 6 50.0 10 18.1 3 33.9 1 M M Gr

K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509


RS83 83 Reichsmuseum, Sweden (S) 2.8 8.6 11.3 M m Bl
HV65A 65 Hofburg Imperial Palace, Vienna, A 4.1 14.0 18.2 M M m t G/t Br
HV65B 65 Hofburg Imperial Palace, Vienna, A 3.6 40.5 44.1 M M Gr
HV65C 65 Hofburg Imperial Palace, Vienna, A 2.2 20.4 22.6 M M C/t Gr
HV65D 65 Hofburg Imperial Palace, Vienna, A 4.5 30.0 34.5 M M Gr
HV85A 85 Hofburg Imperial Palace, Vienna, A 6.4 27.6 34.0 M m t Bl
HV95A 95 Hofburg Imperial Palace, Vienna, A 3.1 10.3 13.4 M M Gr
HV95B 95 Hofburg Imperial Palace, Vienna, A 3.9 34.7 38.6 M M m G/t Br
HV95C 95 Hofburg Imperial Palace, Vienna, A 5.6 13.3 18.9 M m t G, F/t Bl
HV95D 95 Hofburg Imperial Palace, Vienna, A 5.1 22.5 27.7 m m t G/t Bl
D0C , D0B
and D0T
were measured metallographically, and are designated Dco and DBo in Figs. 8 and 9, respectively. DIC
and DIB
were calculated from the ICP
measurements, and are designated Dci and DBi in Figs. 8 and 9, respectively. C = Cuprite Cu2O, B = brochantite, P = posnjakite, A = atacamite, O = other:
G = gypsum, C = carbon, F = Fe2O3, M = major, m = minor, t = trace. Col = Colour, Gr = green, Br = brown, Bl = dark green to black.

2491
2492 K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509

Fig. 5. Thickness data for cuprite and brochantite layers on copper exposed to the Brisbane atmosphere. Data for
up to 2 years was from atmospheric exposure coupons, whereas data for longer exposures came from an analysis
of roof samples.

brochantite growth rate of 0.2 lm/y of the brochantitie layer indicates that 30% of the bro-
chantite is retained and 70% lost as run o.
This estimate is consistent with the weight change data, Table 4 and Fig. 6. The solid
curve in Fig. 6 shows a calculation of the weight gain based on the assumption that all
the oxidised copper had been converted to cuprite. Comparison with the measured weight
gain reveals that there must have been some cuprite loss for exposure times greater than 1
year. Good agreement is obtained between measured and calculated values if it is assumed
that there is a constant removal rate of cuprite 0.2 lm/y and a brochantite deposition rate
of 0.15 lm/y.
The 30% brochantite retention rate can be used to evaluate long term corrosion rates
from the data of patina thicknesses. Fig. 2 shows the resultant plot for Brisbane. It is clear
that there is one trend line through all the data.
Fig. 7 shows a comparison with our data and all existing atmospheric corrosion data
for copper [2032]. As the details of brochantite retentions are not known at the other

Fig. 6. Measured and calculated weight gains for copper exposed to the Brisbane atmosphere at UQ.
K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509 2493

Fig. 7. Comparison of all atmospheric corrosion data.

sites, the corrosion rates for the patinated samples in Fig. 7 have been evaluated on the
basis that all the corrosion products have remained on the surface.

4.3. Formation of a green patina

Fig. 8 shows the plot of the cuprite thickness data plotted against patina age for all
patina samples. This gure shows a trend similar to that shown in Fig. 5 for the patinas
with ages up to about 140 years. For patinas with ages up to 230 years, the cuprite layer

Fig. 8. Comparison of cuprite thickness data for all patinas.


2494 K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509

thickness was generally in the range of 410 lm. For the two Danish (DK) patina samples
of ages 324 and 332 years, the cuprite layer was signicantly thicker at 18 and 19 lm. The
cuprite thickness data for Frederiksberg Castle aged 153 year evaluated from the ICP mea-
surements appears anomalously high compared with the other data and has been disre-
garded in drawing the trendline.
It is worthwhile stressing that the cuprite thickness data shown in Fig. 8 (and the cor-
responding corrosion rates given in Figs. 2 and 7) relate to outdoor exposures. These out-
door cuprite layers are two orders of magnitude thicker than literature data for the
thickness of oxides on copper which was exposed indoors in NJ, USA [36].
Optical and SEM observations of the sample surfaces of the patina samples from Bris-
bane indicated that there is an initial incubation period in which brochantite formed on
the cuprite surface as isolated islands that grew together. During this initial stage the cop-
per looked brown. By about 25 years however, there was a continuous brochantite layer,
which hid the underlying cuprite layer and the samples had a green appearance. Thereaf-
ter, the brochantite layer increased in thickness.
Fig. 9 shows a plot of the brochantite-thickness data for patinas from Brisbane (B),
USA, Denmark (DK), France (F), Sweden (S) and Austria (A). A trend line has been
drawn through the thickness data for patinas from Brisbane Australia indicating linear
brochantite growth kinetics 0.2 lm/y. The data for RUQ43 and SUQ45 fall somewhat
below the trend line, this is attributed to the fact that these samples were from the main
St. Lucia campus, and are more sheltered and are a greater distance from the sea and from
the Brisbane river. The thickness for the brochantite from the USA patina fall within the
scatter for the thickness for the Brisbane patinas. The patinas formed in Denmark, France
and Sweden up to about 150 years in age had a somewhat thicker brochantite layer than
the Brisbane patinas. In fact, the line of best t through the brochantite thickness data for
the patinas from Denmark, France and Sweden up to 100 years of age is a line parallel to

Fig. 9. Comparison of brochantite thickness data for all patinas.


K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509 2495

the line through the Brisbane data. This can be interpreted as indicating a very rapid initial
brochantite growth rate, followed by linear growth kinetics similar to those observed for
Brisbane.
The Danish patinas can be considered to be from two dierent epochs. The oldest sam-
ples (D230, D324 and D332) were originally exposed before the industrial revolution when
there was a very low concentration of SO2 and this correlates with total patina thickness
for these three samples lower than would be expected by an extrapolation from the youn-
ger patinas from Denmark. This agrees with the observation that these old Danish patinas
had cuprite thicknesses greater than the younger patinas.
The samples from the Hofburg Imperial Palace in Vienna exhibited much scatter in the
values of brochantite layer thickness. Many had a brochantite layer comparable in thick-
ness to those from Sweden, Denmark or France, whereas two had a brochantite layer
much lower in thickness. This could correlate with a signicant variety in the patinating
environments experienced by these patinas which also could relate to the patina colours
of some of these patinas.
Characterisation of the patina compounds using XRD revealed cuprite and brochantite
to be the main phases present as detailed in Table 5. For Brisbane and Danish patinas, the
green colour (for the patina as observed with the naked eye) correlated with the presence
of a substantial brochantite layer. Traces of posnjakite were detected in the atmospheric
corrosion samples exposed in Brisbane (Table 1), in the youngest Brisbane patina and
in some of the other patinas (Table 5). Traces of atacamite (Cu2Cl(OH)3) were present
in the majority of Brisbane patinas and some of the other patinas analysed. The presence
of atacamite indicates a signicant local chloride deposition rate in Brisbane the sources of
which is most probably from (1) the Brisbane River, a brackish tidal river, which bisects
the city and (2) the strong local sea breezes. The work of Cole and co-workers [4548] indi-
cates that tidal rivers are not likely to be signicant sources of salt.
Skyward surfaces of HE, HE17 and HE27 were analysed by X-ray photon spectroscopy
(XPS) and the results are presented in Table 6. There was a large surface carbon levels as is
common with XPS measurements. These high surface carbon levels are attributed to
hydrocarbon adsorption on the surface from the atmosphere as controlled XPS experi-
ments by Cousens et al. [37] have shown that atomically clean surfaces pick up substantial
hydrocarbon layers within one minute of exposure to normal laboratory air. To try to
overcome this inevitable carbon contamination, each sample was also analysed after a
5 min argon ion etch at an ion current of 3 lA. The argon ion etch only reduced the

Table 6
XPS analyses of HE7, HE17 and HE27
Sample Condition Atomic %
[C] [O] [Cu] [S] [Cl] [Pb] [Si] [P]
HE7 Before etch 55.2 38.0 4.4 1.0 1.4
After etch 38.4 39.6 9.1 0.8 2.8 1.3 5.4 2.6
HE17 Before etch 64.9 27.5 2.1 0.6 0.3
After etch 61.2 25.6 4.3 0.3 0.9 0.1 6.3 1.3
HE27 Before etch 42.8 45.4 4.3 2.0 0.2 4.3 1.1
After etch 32.3 47.8 7.9 1.9 0.2 8.5 1.4
2496 K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509

carbon concentration slightly. This is attributed to the rough patina surface and the por-
ous nature of the patina. After ion etching, Cu:S ratios for HE7 and HE17 were 11.4 and
14.3, respectively, which is consistent with a surface covered predominantly with cuprite as
was observed by optical microscopy. In contrast, after ion etching, the Cu:S ratio for
HE27 was 4.1 which is consistent with a surface covered predominantly with brochantite
as was observed by optical microscopy. The presence of chloride is consistent with the
XRD detection of traces of atacamite in HE17. The lead is assumed to originate from vehi-
cle emissions, whereas the Si and P is attributed to dirt or sand on the patina surface or as
inclusions within the patina.
Elemental concentrations of Cu, S and Fe per unit area of representative patina samples
from Brisbane, Danish and Swedish patinas, Table 7, were determined by dissolving the
patina from the copper substrate and analysing the solution with ICP. The brochantite
(B) thickness was calculated assuming that all the sulfur was present in the brochantite,
and using a density of qbro = 3.78 g/cm3 [6]. The cuprite layer thickness was calculated
by evaluating the amount of copper in the cuprite layer (total copper less copper in bro-
chantite) and using a density of qcup = 6.00 g/cm3 [6]. These calculated thickness values are
in excellent agreement from those determined by metallography. The mass of the patina
was measured from (mass of sample with patina)  (mass of sample after chemical
removal of patina). The mass of the patina was calculated from the mass of brochantite
plus the mass of cuprite. There was excellent agreement between these two values for
the patina mass providing additional support for the self consistency of these evaluations.
The nal column gives the amount of Fe (as measured by ICP) as a percent of the copper
present in the brochantite.
Fig. 10 presents a plot of measured surface concentration of sulfur, [S] as a function of
measured brochantite cross-sectional thickness, x (lm). The lines correspond to various
degrees of porosity. The line denoted by 0% porosity is represented by the equation
S mg=cm2 x lm kqbro =M bro =nS M S 8
where Mbro is the molecular weight brochantite, nS is the number of sulphur atoms in
brochantite, MS is the molecular weight sulphur, qbro is the density of pure brochantite

Table 7
ICP measurements of atomic concentrations of Cu, Fe and S per unit area of patina
Sample Elemental concentration Thickness Patina mass [Fe]/[Cu]
(mg/cm2) (lm) (mg/cm2)
[Cu] [Fe] [S] B Cuprite measured calculated wt% in B
RUQ43 4.84 0.018 0.06 2.5 8.1 5.8 3.49
SUQ45 4.91 0.018 0.04 1.6 8.6 8.9 5.8 8.87
D26 6.15 0.021 0.35 13.2 6.3 9.1 8.8 0.76
D47 7.90 0.044 0.37 13.7 9.4 9.9 10.8 1.51
D55 10.17 0.018 0.37 13.8 13.6 13.1 13.4 0.61
D70 8.10 0.027 0.53 19.7 7.4 12.3 11.9 0.64
D82 11.39 0.010 0.51 18.9 13.9 14.9 14.9 0.26
D99 13.13 0.093 0.59 22.0 18.6 20.1 17.8 1.99
CRS150 26.10 0.076 0.84 31.2 36.5 33.4 33.7 1.15
D230 10.57 0.090 0.63 23.4 10.5 14.8 15.1 1.77
D324 18.43 0.019 0.68 25.3 24.6 27.2 24.3 0.37
D332 16.90 0.024 0.91 33.9 18.1 25.6 23.7 0.34
K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509 2497

Fig. 10. Measured surface sulphur concentration plotted as a function of measured brochantite cross-section
thickness. The lines correspond to various degrees of porosity.

mineral, k = 0.1 (units conversion factor). The lines indicate 0%, 25% and 50% porosity.
Vertical error bars represent the error associated with the ICP analysis technique and are
predominately due to variations in patina thickness within a particular copper roof sam-
pled. Horizontal error arises also for variations in patina thickness observed from cross-sec-
tions. Due to the signicant errors, limited information can be obtained from this
technique. Nevertheless, this analysis indicates that there is signicant porosity for the Dan-
ish patinas aged between 26 and 99 years, consistent with the way they absorb water [9,10].

4.4. Rust coloured (Brown) patinas

Concentrations of Fe in the patinas varied from 0.3% to 8.9% of the Cu in the brochan-
tite, Table 7. The Queensland patinas RUQ43 and SUQ45 had the highest Fe concentra-
tion and these had a brownish tinge. XRD had not revealed any separate Fe phase. To
measure the solubility of Fe in the brochantite crystal lattice, brochantite was precipitated
by adding NaOH to solutions of copper/iron sulphate having [Fe]/[Cu] ratios of 0.79, 1.58
and 3.16 wt%. The [Fe]/[Cu] ratio in the brochantite produced was 0.69, 1.37 and
2.03 wt%. Brochantite precipitated from pure copper sulphate (with no Fe) appeared
blue-green. The brochantite became greener with increasing Fe content, with a Fe content
of 0.692.03 corresponding to the colour of natural patinas. XRD analysis of brochantite
containing 0.692.03 wt% showed brochantite to be the only phase present, and there was
no measurable change in peak positions, intensities or widths, Fig. 11. This indicates that
the Fe is in solid solution within the brochantite crystal structure. This correlates with the
similar ionic radius for Fe2+ and Cu2+ at 7.7 and 7.1 nm, respectively, although there is
sucient dierence in the electronic conguration to produce colour changes.
2498 K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509

Fig. 11. XRD scans of Brochantite containing [Fe]/[Cu] = 0.69%, 1.37% and 2.03%.

4.5. Black and brown patinas

The surface densities of iron and sulphur as determined by ICP-AES for selected Aus-
trian patina samples and the dark green to black Swedish patina sample are shown in
Table 8. The surface density of sulphur was used to calculate a surface density of copper
present in the outer patina, assuming that all the sulphur is bound up in brochantite. The
measured surface iron content was then used to determine the [Fe]/[Cu] ratio in the outer
patina layer, and these are the values in the last column of Table 8.
Line proles of copper and iron (background subtracted) for various Hofburg Imperial
Palace patinas are shown in Fig. 12. Patinas HV65A (Fig. 12(a)) and HV95B (Fig. 12(b))
have signicant amounts of iron in the very outermost patina layers. In contrast HV85
(Fig. 12(c)) contains little iron. The line marked A in Fig. 12 represents the approximate
boundary between the copper substrate and the cuprite layer of the patina.
Several trends in patina colour and iron content became apparent. When brochantite
was the predominant phase, as shown by XRD and the [Fe]/[Cu] ratio was low, such as
the sample AFC117, then the patina was green in colour. If however the [Fe]/[Cu] ratio
was high approaching 10%, then the patina was brown as for example samples HV65A
and HV95B. Moreover, EMPA line proles (Fig. 12) indicated that the iron was concen-
trated in the top most layers. This is consistent with the observations at the Customs
House in Brisbane. Eyewitness reports indicate that this 150 year old copper dome had
changed from a pleasant green patina to have a rust coloured appearance over the last
2040 years, and moreover, the extensive restoration eort, which involved removal of
the outermost layers, has restored the dome to a more pleasing green colour.

Table 8
ICP measurements of atomic concentrations of Fe and S per unit area of patina
Sample Elemental concentration (mg/cm2) [Fe]/[Cu]
[S] [Fe] [Cu] at% in B wt% in B
RS83 0.8 0.0 6.2 0.0 0.0
HV65A 0.5 0.3 4.0 9.7 8.5
HV95B 0.5 0.4 4.3 10.2 8.9
HV95D 0.2 0.0 1.4 1.5 1.3
HV85 0.2 0.1 1.6 3.5 3.1
HV95C 0.2 0.1 1.8 5.8 5.1
K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509 2499

Fig. 12. Line proles of copper and iron (background subtracted) across Hofburg Imperial Palace patina cross-
sections: (a) HV65A; (b) HV95B; (c) HV85.

If the outer layer was a minor constituent of the patina, and cuprite was present, then
the samples appeared dark green to black (designated as black in Table 5). These samples
(HV85, HV95C, HV95D and RS83) had [Fe]/[Cu] ratio values ranging from 0% to 3.5%.
The presence of cuprite as the major constituent is expected to account for the dark col-
ours observed.

5. Discussion

5.1. Atmospheric conditions

The UQ site in Brisbane, Australia is about 15 km from the coastline and is character-
ised by a rainfall of around 1000 mm per year, average relative humidity between 50% and
80%, and temperatures between 10 and 30 C. Even though Brisbane is relatively near the
coast, the structures considered in this paper were all more than 15 km distant from the
coast. That far inland the main sources of SOx are considered to be (1) atmospheric
SO2 and (2) SO2 4 from rain. Meteorological data for the period 19901994 was examined
2500 K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509

to evaluate the time of wetness, tw, dened as the fraction of time that the relative humidity
was equal to or above 85%: tw = 2165 h/8766 h = 0.247. At least 80% of the time of wet-
ness occurred between 8 pm and 8 am. The average number of wet days per year was 106
which is equivalent to a rain event every 3.5 days. No detailed analysis of Brisbane rain-
water was available. The closest studies relate to Sydney and Melbourne. The average pH
of Sydney rainwater was 4.6 [38]. Rainwater in Melbourne varied between 4.4 and 6.6 in
pH with only 18% of all rainfall events samples being less than 5.0 in pH [39]. Several sam-
ples of rainwater were collected at UQ during the atmospheric exposure tests. The average
pH recorded was 5.5; no samples had a pH less than 5. Rainwater analysed by high per-

formance liquid chromatography indicated 35 lM SO2 4 and 118 lM Cl . There was some
industrial pollution with a 0.002 ppm SO2 level corresponding to a SO2 deposition rate of
50 mg SO2/(m2 day) according to ASTM G91-86. This corresponds to the Moderate-
Urban/Light industrial environmental category outlined in AS 2312. This environment
is expected to result in low corrosion rates. This corresponds to the corrosion rates mea-
sured which showed an average corrosion rate at two years of about 0.9 lm/y. The dier-
ent specimen materials (TCU, ESU and EAU, all 99.9% Cu) initially showed dierent
corrosion rates which was attributed to slight dierences of surface condition, particularly
surface roughness. After three months of exposure, all the copper materials showed similar
corrosion rates as the protective cuprite layer had formed [1].

5.2. Skyward and groundward surfaces

Brochantite and posnjakite crystals were observed on the copper surfaces exposed at
Brisbane after 6 m, 1 y, and 2 y on both the skyward and ground ward surfaces. There
was about 20% more brochantite on the skyward surfaces and the data from these atmo-
spheric exposures is consistent with the data from the skyward surfaces of the natural
patina samples. The exposure racks used in this work were open allowing relative easy
atmospheric access to both sides of each specimen and moreover even the ground ward
surfaces would often be wet during rain events, particularly when these were associated
with even moderate winds, as was often the case. This is very dierent to the typical situ-
ation for copper exposed on a roof where the underside is dry and not wet by rain. Indeed,
it is often a prime function of a roof to keep the underside dry. The ground ward side of
the patina samples was typically copper coloured, with the colour ranging from copper
brown to brown to black with increasing patina age and did not show any green patina
compounds even after 320 year exposure.

5.3. Atmospheric corrosion

The XRD and SEM results of the coupons exposed at Brisbane indicated that cuprite
was the initial corrosion product and that brochantite formed subsequently on the cuprite
surface as isolated crystals. After 1 year exposure there was a small amount of posnjakite.
Posnjakite is a hydrated form of brochanitite and has been previously identied in short-
term exposures. Copper roof samples from the Brisbane environment examined by XRD
revealed a trace posnjakite in only the earliest sample (7 year exposure) and increasing
amounts of brochantite with increasing age. The posnjakite formed initially either dis-
solved or was converted to brochantite. XRD analysis of Brisbane patinas indicated that
there were only minor amounts of the copper hydroxychloride, atacamite.
K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509 2501

ICP analysis of corrosion products for the TCU samples exposed at UQ for 2 years
indicated the following average picture, see Tables 24. There had been an average copper
weight loss, DWCu = 0.76 mg/cm2/y, which had been converted to 0.80 mg cuprite/cm2/y.
Of this cuprite, a mass DWCu_oxidised = 0.2 0.1 mg/cm2/y had been further oxidised to
Cu2+, of which W Cu bro 0:04  0:02 mg/cm2/y remained on the surface as brochantite
and W Cu runoff 0:14  0:08 mg/cm2/y was the amount of runo. The amount of copper
in the runo could be described in the following two ways, (1) as 70% of the mass of cop-
per as cuprite further oxidised to Cu2+ or (2) as 18% of the metallic copper lost from the
sample.
FitzGerald et al. [6], using titrations of copper sulphate solutions with sodium hydrox-
ide, showed that brochantite and posnjakite nucleate from solution in a matter of minutes
even at the very low concentrations of Cu2+ and SO2 4 . Therefore the Cu
2+
lost as runo is
2+
either as Cu ions or as brochantite precipitates that have not adhered suciently to the
surface of the cuprite. Optical microscopy of young patinas revealed that brochantite crys-
tals formed conglomerates growing from a common nucleation point. This was also evi-
dent from the experiments of FitzGerald et al. [6] involving articial rainwater and
cuprite powder. Filtering of brochantite and posjnakite precipitates from solutions of cop-
per sulphate produced a coherent lter cake when the precipitates were not washed thor-
oughly with alcohol. This suggested that the adherence of these brochantite conglomerates
to the cuprite layer or to themselves is increased by the drying of the moisture layer. A
secondary precipitation of brochantite crystals bond the larger crystals together. The sec-
ondary precipitation of brochantite between the already present precipitates is a result of
the increased Cu2+ and SO2 4 concentrations above the solubility limits, i.e. 23 ppm, due
to evaporation of the moisture lm. A schematic diagram is presented in Fig. 13.

5.4. Mechanism of patination

The observations of the formation of the corrosion products, e.g. see Fig. 1 can be sche-
matically illustrated as in Fig. 13. This has led to the following model for patina forma-
tion, Fig. 14. Cuprite forms on the copper surface by the oxidation of copper by
reaction (R1)
2Cu 0:5O2 Cu2 O R1

Fig. 13. Schematic illustration of the form of brochantite crystals on the cuprite surface.
2502 K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509

Fig. 14. Illustration of the reactions involved in brochantite formation during patination.

The oxidation rate of (R1) is controlled by the diusion of Cu+ through the cuprite layer
and is equivalent to the dissolution/oxidation rate of cuprite (R2) after approximately 70
years in Brisbane whereupon a steady state corrosion rate is reached. Surface copper ions
from the cuprite are further oxidised from Cu+ to Cu++ in the presence of a layer of sur-
face water via the two partial electrochemical reactions (R2)
Cu Cu e
R2
0:5O2 H2 O 2e 2OH
Simultaneously atmospheric SO2 is oxidised to sulphate cations by reaction (R3).
SO2 H2 O 0:5O2 H2 SO4 R3
Some sulphates can be supplied directly to the cuprite surface by rain, but the supply of
sulphates to the surface is not sucient to account for the quantity of brochantite present
in Brisbane patinas. Brochantite is precipitated by reaction (R4) when the pH in the water
becomes suciently alkaline, as determined by the brochantite stability domain in the
Pourbaix E-pH diagram [9]
4Cu2 SO 
4 6OH ! CuSO4 3CuOH2 R4
Brochantite thus precipitates from an aqueous surface layer from an oxidative dissolution
of some cuprite. Not all the brochantite is retained on the copper surface. Some is lost in
run o and this leads to the commonly observed staining of the base of copper roofs and
bronze statues.
The amount of Cu2+ lost to runo increases as the acidity of the washing rain increases
[33]. Copper lost from the surface occurs by the dissolution of cuprite and brochantite and
the carrying away of Cu2+ or ne brochantite crystals by owing water. The production of
brochantite occurs in the stagnant lm after the rain shower and in an absorbed surface
water lm and furthermore adherence to the cuprite is enhanced by evaporation of this
water lm. It therefore follows that surfaces experiencing long rain showers compared
K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509 2503

with several short showers of the same pH would loose more copper to runo and form
less brochantite.
The incorporated sulphur surface density from Fig. 9 for Brisbane patinas was
5 103 mg/(cm2 y). By comparison, the atmospheric SO2 concentration corresponds to
a SO2 deposition rate of the order of 0.5 mg S/(cm2 y) which is 100 times greater than
the amount incorporated in the patina. Since brochantite growth only occurs when the
surface is covered by a moisture lm, the actual brochantite growth rate can be expressed
2
as 0.8 0.2 lm/(y tw). This requires a supply of SO2 4 of 2 10 mg S/(cm2 y) which is
still 2040 times smaller than the deposition rate. Sulphate production via the oxidation
of SO2 to SO2 4 is pseudo rst order with respect to SO2 having a rate constant of
3 103 s1 [40]. This is much greater than the sulphate consumption by the formation
of brochantite and therefore poses no constraint to brochantite formation. In addition
to the supply of sulphate by the oxidation of SO2, sulphates are already present in the rain-
water. A sulphate concentration of 35 lM was measured in Brisbane rainwater. This was
below the solubility limits of posnjakite at the pH of Brisbane rainwater of 5.5. The sol-
ubility limits of posnjakite had a pH minimum solubility of 6.6 with aCu2
aSO2 41:4 lM. However the oxygen reduction reaction (R2) would increase the pH of
4
the water lm and evaporation of the water lm would precipitate posnjakite and/or bro-
chantite. Consider that a cuprite surface experiences 106 rainy days per year and the stag-
nate water lm is 100 lm thick. Then a minimum sulphate supply from the evaporation of
the water lm is approximately 1 103 mg S/(cm2 y). This is a fth of the total sulphur
incorporated into the patina. The average number of rain events per rainy day is most
likely to be less than ve and therefore the quantity of preoxidised sulphates already pres-
ent in rainwater prior to wetting the surface is not sucient to account for the quantity
found in Brisbane patinas. This indicates that the majority of sulphates found in Brisbane
patinas have been supplied by the adsorption and oxidation of SO2.
Since cuprite oxidation occurs only when the surface is covered by a moisture lm the
actual oxidation rate per tw was 0.65 0.3 mg/(cm2 y tw). Considering a moisture layer on
average is less than 100 lm, then its concentration would be at least 100 ppm Cu2+ within
a minute. This indicates that brochantite and/or posnjakite precipitate from solution
within minutes of a surface adsorbing a moisture layer.
Seventy percent or 0.14 0.08 mg Cu2+/(cm2 y) of cuprite oxidised was lost as runo.
This quantity is a physical characteristic of the exposure environment being inuenced by
the pH of the rain, the frequency of rain showers and the duration of these showers. The
supply of sulphate into the water lm is sucient with a supply 2040 times greater than
that being incorporated into the patina layer. It is therefore deduced that the rate limiting
step in the growth of the brochantite layer is the oxidative dissolution of cuprite to Cu2+.
This oxidative dissolution is inuenced by the pH of the rain, which is also inuenced by
the SO2 concentration amongst other variables.
The mass balance of total copper corroded can be represented by
P t P cup P bro P runoff 9
where Pt is the total penetration depth (lm), Pcup is the that part of total penetration depth
(lm) attributable to the cuprite layer, Pbro is the that part of total penetration depth (lm)
attributable to the brochantite layer, and Pruno is the that part of total penetration depth
(lm) attributable to the runo. The growth of the cuprite layer can therefore be repre-
sented by
2504 K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509

P cup ktn  P bro P runoff 10


where k and n are the constants of the bi-logarithmic corrosion law, Pbro = 0.045 lm (of
Cu)/y, and Pruno = 0.15 lm/y. Eq. (10) has been tted to the measured cuprite equivalent
thickness data in Fig. 15 generating the values for the constants n = 0.71 and k = 0.84 lm/y
in good agreement with the n and k values evaluated from Fig. 2. As the cuprite layer thick-
ens, the diusion of Cu+ cations through the oxide becomes more dicult, consequently
decreasing the growth rate. The oxidation rate of cuprite, assuming the eects of brochan-
tite layer are considered negligible is a function of the environment involving parameters
such as rain acidity, time of wetness, frequency and duration of rain events, SO2 deposition
rates and temperatures. The growth rate of the cuprite layer is represented by
dP cup =dt nktn1  dP bro  P runoff =dt 11
Substituting dPcup/dt = 0, n = 0.71, k = 0.84 mm/y, and d(Pbro  Pruno)/dt = 0.18 lm
Cu/y into Eq. (11), a maximum cuprite thickness of 5.2 lm is obtained after 70 years
in good agreement with the observed trend.
The bi-logarithmic nature of the atmospheric corrosion of copper is attributed to the
protective nature of the cuprite. However, when the dissolution or quantity of this lm lost
to runo is signicant in comparison to the total amount of copper corroded, the corro-
sion products are less protective corresponding to higher n values. From Eq. (11) a steady
state corrosion rate is reached after 70 years, 0.18 0.08 lm Cu/y, where it remains until
environmental factors increase or decrease the rate of copper lost to runo. When consid-
ering copper corrosion over 70 years, the copper lost to runo eects the average corrosion
rate signicantly. Average corrosion rates would tend to asymptotically approach the

Fig. 15. Curve tted equivalent thicknesses for the mass balance of corroded copper.
K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509 2505

steady state runo rate. This explains the large n value derived from Eq. (11), which was
based on 142 years of copper exposure.
Desirable conditions for brochantite formation are: the presence of a stagnant water
lm after rain or an adsorbed moisture lm from high relative humidity, a large time of
wetness, no (or little) runo water, a low rainwater pH, and a sucient supply of sulphate
ions. Undesirable conditions are: long rain showers and large amounts of runo water, a
dry environment, a very quick drying time limiting the time a stagnant water lm is present
and a low rainwater pH. The pH of the rainwater can have both positive and negative
eects on brochantite growth. Low pH enhances brochantite layer formation when the
low pH of the rainfall promotes oxidative dissolution of the cuprite and the subsequent
precipitation of brochantite if the environment is humid, has frequent short rain showers
with little runo, the stagnant water lms is present for long periods of time and drying of
the stagnant water lm precipitates brochantite to consolidate the existing brochantite on
the surface. In contrast, low pH will dissolve both cuprite and previously formed brochan-
tite under environmental conditions where there are long rain showers with most of the
surface water owing o the surface and the chemical compounds lost in the runo.
The work of Kucera and Collins [33] indicated that the amount of copper lost in runo
in Sweden increased with increasing acidity (with pH values decreasing from 4.3).

5.5. Green patinas

Copper patinas, formed by exposure of copper to the atmosphere, posses two distinct
layers. The inner layer always consisted of cuprite with copper in the +1 oxidation state.
The outer layer was most often composed of brochantite. Posnjakite, which is the hydrate
form of brochantite, was detected in the 7 year old Brisbane patina. Traces of atacamite
were detected in most of the Brisbane patinas because of the local sources of chlorides. The
oxidation sate of copper was +2 in the outer patina layer, for all cases including brochan-
tite, posjnakite and atacamite.
Optical metallography showed the presence of green crystal conglomerates after 10 days
of exposure on both the skyward and groundward surfaces of the exposure coupons. As
the exposure time increased the surface density of these green conglomerates also
increased. It is evident that posjnakite and/or brochantite was present on the surfaces
of the exposure samples after 10 days but their surface densities were below the detection
limits of XRD. After 1 year of exposure posjnakite and brochantite were present in su-
cient quantities to be detected by XRD.
Chlorides were detected by XPS on all exposure coupons and their surfaces except the
groundward surface after 10 days of exposure. This was consistent with the EDS analysis.
However XRD scans did not show the presence of chloride phases. The chloride is
expected to have been present as a copper chloride namely atacamite but below the
XRD detection limit. This is consistent with the results of the XRD analyses of the Bris-
bane patinas which indicated minor amounts of atacamite.
Brisbane patinas take from 17 to 27 years to change from the brown/black colour char-
acteristic of cuprite to the green of brochantite. There is green brochantite on the surface
before 27 years, but the brochantite is in isolated crystallite conglomerates, which do not
have sucient hiding power to cover the underlying brown/black cuprite. The patina col-
our changes to green when the islands of brochantite nally meet to form a continuous
brochantite layer of sucient thickness (or the order of 5 lm).
2506 K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509

For a two layer structure as observed for these patinas, it is most likely that the rst
formed layer is the protective layer, that is the corrosion protection is assigned to the cupr-
ite layer next to the copper metal. This is consistent with the porous nature of the brochan-
tite layer which forms on top of the cuprite layer.
The protective qualities of the patina requires an assessment of each of the two patina
layers. Cuprite Cu2O is a p-type semi-conductor. It has an intrinsic non-stoichiometric
composition, which is more accurately represented as Cu2xO where x is a small fraction.
During oxidation, cation vacancies (VCu+) are formed at the outer oxide surface and these
diuse through the cuprite lattice towards the metal surface, which is equivalent to the
migration of Cu+ cations from the metal/oxide interface to the oxide surface [41,42].
The model of atmospheric corrosion indicates that the brochantite layer grows by the
oxidative dissolution of the cuprite layer and the precipitation of brochantite crystallites,
which are incorporated into the upper patina layer. The protective quality of the cuprite
layer can be determined from the gradient of the plot of log cuprite layer thickness versus
log time. A line of best t drawn through the rst 17 year data where the eects of the bro-
chantite layer can be considered to be negligible had a gradient or n value of 0.51 and a
correlation coecient of 0.98 indicating a protective layer. A limiting cuprite thickness
was reached after 27 years: then the rate of formation was equal to the rate of dissolu-
tion. The brochantite growth rate followed a linear function which is equivalent to an n
value of 1, indicating a non-protective layer.
The linear growth rate of brochantite for Brisbane patinas was 0.2 0.05 lm/y, see
Fig. 9. The brochantite growth rate appeared linear for at least the rst 142 years. It is
possible that older brochantite layers might have much lower porosity and then might
begin to hinder transport of Cu2+ and O2 through the brochantite layer and consequently
reduce the corrosion rate further. Experimental support for this expectation has been pro-
vided by the electrochemical studies of Leygraf and co-workers [4953].

5.6. Atmospheric corrosion rate and the corrosion rate indoors

It is worthwhile stressing that the total patina thickness shown in Fig. 9 (and the cor-
responding corrosion rates given in Figs. 2 and 7) relate to outdoor exposures. These out-
door patinas are two orders of magnitude thicker than literature data for oxides on copper
which was exposed indoors in NJ, USA [36]. This indicates that the outdoor environment
degrades the protective nature of the cuprite layer by two orders of magnitude. This could
be caused by the interaction of a water lm on the surface of the cuprite, which might
accelerate ion transport through the cuprite lm by a electric eld mechanism. An alter-
native might be suggested by the cracks evident in the cuprite lm of Fig. 1 and also in
[49,52]. Such cracks if present during corrosion would provide pathways for rapid
reactions.

5.7. Measured corrosion rate and prior understanding

These studies of the atmospheric corrosion of copper have shown that the patina on
pure copper is protective and that the patina causes a continuing decrease in the corrosion
rate of copper with increasing exposure time. An atmospheric corrosion rate of 1 lm/y is
typical for copper exposed to a normal atmosphere for a period of about 10 years. The
corrosion rates measured in this work were comparable to those measured in the literature.
K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509 2507

Moreover, the current research has indicated that it is possible to estimate the atmospheric
corrosion rates of copper over long period of time from a detailed examination of copper
samples that have been exposed to the atmosphere as roofs for decades and centuries.
These results have indicated that it is possible to map out long term atmospheric corrosion
rates using an analysis of copper patinas.

5.8. Patina colour

The examination of the various natural patina samples showed that natural patinas can
come in a variety of colours which corresponds to observations of natural patinas
observed when travelling in Australia, Europe and the USA. The colour is controlled
by the amount of the patina and its chemical composition. Thin patinas containing pre-
dominantly cuprite were black. If the patina was suciently thick, and the [Fe]/[Cu] ratio
was low, then the patina was blue to green, whereas if the [Fe]/[Cu] ratio was approxi-
mately 10 at%, then the patina was brown in colour. There were no discrete iron phases
present. Rather, the iron was in solid solution in the brochantite, which might be desig-
nated as a (copper/iron) hydroxysulphate. In the brown patinas examined, the iron was
distributed predominately in the outermost part of the patina.

5.9. Articial patinas

The formation of a patina on external architectural copper is often considered to be


aesthetically pleasing and many builders and architects desire this surface colouration from
the outset from the time of rst completion of the building or structure. To achieve this
immediate eect, we developed CHP, a patination process to produce a patina, which
resembles the natural patina in all ways [43]. CHP applied to the copper surface a micro-por-
ous layer of the natural patina compound bound together in a porous array with a commer-
cial inorganic silicate binder to have a structure similar to that of a natural patina as
illustrated in Fig. 4. The CHP patina compound matched (chemical composition, colour,
morphology) the main compound (brochantite) that is found in natural patinas. A process
was developed to produce this natural compound in the appropriate form. A micro-porous
layer was produced which would allow natural patination to occur in the interstices. CHP
could be applied by traditional coating application methods (e.g. as a spray or with a brush)
[44].

6. Conclusions

During the atmospheric corrosion of copper, a two-layer patina forms on the copper
surface. Cuprite is the initial corrosion product and cuprite is always the patina layer
in contact with the copper. The growth laws describing patina formation indicate that
the decreasing corrosion rate with increasing exposure time is due to the cuprite layer.
The corrosion rates in atmospheric exposures outdoors being two orders of magnitude
greater than corrosion rates indoors, indicate that the protective nature of the cuprite
lm is degraded by the presence of a water layer on its surface. Green patinas are typ-
ically characterised by an outer layer of brochantite, which forms as individual crystals
on the surface of the cuprite layer, probably by a precipitation reaction from an aque-
ous surface layer on the cuprite layer.
2508 K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509

Natural patinas can come in a variety of colours. The colour is controlled by the
amount of the patina and its chemical composition. Thin patinas containing predomi-
nantly cuprite are black. If the patina is suciently thick, and the [Fe]/[Cu] ratio is low,
then the patina is blue or green, whereas if the [Fe]/[Cu] ratio was approximately
10 at%, then the patina is brown in colour. The iron is in solid solution in the brochan-
tite, which might be designated as a (copper/iron) hydroxysulphate. In the brown pati-
nas examined, the iron was distributed predominately in the outermost part of the
patina.
An articial patina has been developed based on the knowledge of natural patination.

Acknowledgements

This research has been supported by the Australian Research Council, The Interna-
tional Copper Association of NY, Copper Reneries Pty Ltd. and Copperform Pty Ltd.
Thanks are expressed to the many colleagues who supplied natural patina samples.

References

[1] J. Nairn, R.G. Skennerton, A. Atrens, Journal of Materials Science 38 (2003) 995.
[2] A. Atrens, in: R.N. Parkins (Ed.), Life Prediction of Corrodible Structures, NACE, Cambridge, 1991.
[3] J. Nairn, K. FitzGerald, A. Atrens, in: International Conference on Metals Conservation, Metal-95, Semur
en Auxois, France, 1995, p. 86.
[4] A. Atrens, K. FitzGerald, G. Skenerton, J. Nairn, in: 13 International Corrosion Congress, Melbourne,
Australasian Corrosion Association, 1996, paper 42.
[5] J. Nairn, A. AtrensMaterials Research 96, vol. III, The Institute of Metals and Materials, Australia, 1996,
p. 16.
[6] K.P. FitzGerald, A. Atrens, J. Nairn, Corrosion Science 40 (1998) 2029.
[7] K. FitzGerald, Ph.D. thesis, The University of Queensland, 1996.
[8] G. Skennerton, Ph.D. thesis, The University of Queensland, 1998.
[9] T.E. Graedel, K. Nassau, J.P. Franey, Corrosion Science 27 (1987) 639.
[10] J.P. Franey, M.E. Davis, Corrosion Science 27 (1987) 659.
[11] R. Holm, E. Mattson, ASTM STP 767 (1982) 85.
[12] M.K. Kalish, Metallurgiya (1971).
[13] E. Niskanen, U. Maritas, Canadian Metallurgical Quarterly 9 (1970) 339.
[14] H. Abe, Y. Ishi, H. Kato, Quarterly Reports, Railway Technical Testing Institute 12 (1971) 170.
[15] K. Nassau, A.E. Miller, T.E. Graedel, Corrosion Science 27 (1987) 716.
[16] M. Suzuki, H. Koyama, Journal of Metal Finishing Society of Japan 24 (1973) 668.
[17] R.J. Matyi, R. Baboian, Powder Diraction 1 (1986) 299.
[18] A. Salnik, W. Faubel, H. Klewe-Nebenius, A. Wendl, H.J. Ache, Corrosion Science 37 (1995) 741.
[19] A. Vendl, B. Pichler, J. Weber, R. Erlach, Wiener Berichte uber Naturwissenschaft in der Kunst (1991).
[20] L.P. Costas, ASTM STP 767 (1982) 106.
[21] I.R. Scholes, W.R. Jacob, in: International Symposium on Copper and its Alloys, Amsterdam, 1970.
[22] A.W. Tracy, ASTM STP 175 (1956) 67.
[23] E. Mattsson, R. Holm, ASTM STP 435 (1968) 187.
[24] A.P. Costillo, J.M. Popplewell, ASTM STP 767 (1982) 60.
[25] C.R. Southwell, J.D. Bultman, in: Proceedings of the Conference of Atmospheric Corrosion, Hollywood
1980, 1982.
[26] V. Kucera, D. Knotkova, J. Gullman, P. Holler, in: 10th International Congress on Metallic Corrosion,
India 1987, Key Engineering Materials 2028 (1987) 167.
[27] M. Morcillo, S. Feliu, S. Gimenez, in: 10th International Congress on Metallic Corrosion, India 1987, Key
Engineering Materials 2028 (1988) 17.
K.P. FitzGerald et al. / Corrosion Science 48 (2006) 24802509 2509

[28] A. Atrens, Final Report, ICA Project No. 391, International Copper Association NY, 1991.
[29] S. Dean, ASTM STP 1239 (1995) 56.
[30] S. Feliu, M. Morcillo, Atmospheric Corrosion (1982) 913.
[31] M. Morcillo, J. Simancas, S. Feliu, ASTM STP 1239 (1995) 195.
[32] D. Thompson, A. Tracy, J. Freeman, ASTM STP 175 (1956) 77.
[33] V. Kucera, M. Collins, in: Proceedings of the 6th European Congress Metallic Corrosion, 1977, p. 189.
[34] W. He, Licentiate thesis, Royal Institute of Technology Stockholm, Sweden, 2000.
[35] M. Pourbaix, in: W. Ailor (Ed.), Atmospheric Corrosion, Wiley, NY, 1982, p. 107.
[36] R. Schubert, S.M. DEgidio, Corrosion Science 30 (1990) 999.
[37] D.R. Cousens, B.J. Wood, J.Q. Wang, A. Atrens, Surface and Interface Analysis 29 (2000) 23.
[38] G.P. Ayers, R.W. Gilbert, U. Cernot, Clean Air Australia 21 (1987) 68.
[39] G.P. Ayers, R.W. Gilbert, Journal of Atmospheric Chemistry 2 (1984) 25.
[40] J.M. Miller, R.G. Depena, Journal of Geophysical Research 77 (1972) 5905.
[41] H.H. Uhlig, W.R. Revie, Corrosion and Corrosion Control, Wiley, NY, 1985.
[42] J.C. Scully, The Fundamental of Corrosion, Pergamon, 1990.
[43] A. Atrens, J. Nairn, Articial patina, PCT Patent Application No. PCT/AU95/00236, Dated 21 April 1995.
[44] A. Atrens, J. Nairn, H. Fernee, K. FitzGerald, G. Skennerton, A. Olonjana, Materials Forum 21 (1997) 57.
[45] I.S. Cole, D. Lau, D.A. Paterson, Corrosion Engineering Science and Technology 39 (2004) 209.
[46] I.S. Cole, D.A. Paterson, Corrosion Engineering Science and Technology 39 (2004) 125.
[47] I.S. Cole, W.Y. Chan, G.S. Trinidad, D.A. Paterson, Corrosion Engineering Science and Technology 39
(2004) 89.
[48] I.S. Cole, W.D. Ganther, D.A. Paterson, G.A. King, S.A. Furnman, D. Lau, Corrosion Engineering Science
and Technology 38 (2004) 259.
[49] X. Zhang, W. He, I. Odnevall, J. Pan, C. Leygraf, Corrosion Science 44 (2002) 2131.
[50] A. Kratschmer, I. Odenevall, C. Leygraf, Corrosion Science 44 (2002) 425.
[51] I. Odnevall, C. Leygraf, Corrosion Science 43 (2001) 2379.
[52] W. He, I. Odnevall, C. Leygraf, Corrosion Science 43 (2001) 127.
[53] I. Odnevall, P. Verbiest, W. He, C. Leygraf, Corrosion Science 42 (2000) 1471.
[54] A. Salnick, W. Faubel, H. Klewe-Nebenius, A. Vendl, H.-J. Ache, Corrosion Science 37 (1995) 741.
[55] L. Nunez, E. Reguera, F. Corvo, E. Gonzalez, C. Vazquez, Corrosion Science 47 (2005) 461.
[56] I.T.E. Fonseca, R. Picciochi, M.H. Mendonca, A.C. Ramos, Corrosion Science 46 (2004) 547.

You might also like