Engineering Failure Analysis 64 (2016) 111–125

Contents lists available at ScienceDirect

Engineering Failure Analysis

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

Analysis of pitting corrosion failure of copper tubes in an

apartment fire sprinkler system
Sang Hee Suh a,⁎, Youngjoon Suh a, Hyung Goog Yoon b, Jeong Han Oh b, Youngjun Kim c,
KiMin Jung c, HyukSang Kwon c
a
Center for Electronic Materials, Korea Institute of Science and Technology, 5 Hwarang-ro 14-gil, Seongbuk-gu, Seoul 02792, Republic of Korea
b
SH Corporation, 621 Gaepo-ro, Gangnam-gu, Seoul 06336, Republic of Korea
c
Dept. of Materials Science & Eng., KAIST, 291 Daehak-ro, Yuseong-gu, Daejeon 34141, Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history: Microscopical and structural analyses as well as visual inspection of copper tubes were used to
Received 15 December 2015 investigate the cause of pitting corrosion failure of copper tubes in the wet pipe sprinkler sys-
Received in revised form 25 February 2016 tem. Chemical analysis of the water in the copper tubing and XRF/XRD analysis of its sediments
Accepted 7 March 2016
were also used to obtain hints on what was happening in the copper tubing during the prog-
Available online 10 March 2016
ress of the pitting corrosion. It was found from the failed copper tube that a significant amount
of pressurized air was present over the water in the copper tubing during operation and a se-
ries of corrosion pits were aligned adjacent to the water/air line. The waterline localized corro-
sion, a type of differential oxygen concentration cell corrosion, induced by pressurized air over
the water in the copper tubing was identified as the cause of the pitting corrosion failure. A
state of a very low oxygen concentration was maintained under the envelope of a dense
layer of malachite, the corrosion byproduct, which was evidenced by the formation of Cu2O
crystalline particles inside the corrosion pit. CuO particles observed on the inside surface of
the copper tube do not seem to help prevent local as well as general corrosion. Accelerated
pitting corrosion of the copper tube in the wet fire sprinkler system was simulated using a
differential aeration cell. Finally, measures for stopping or delaying the pitting corrosion of
the copper tube in the wet sprinkler system are suggested.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction

Fire sprinkler systems are now extensively used worldwide at commercial buildings as well as homes, schools and apartment
buildings for fire protection. In 2011 Wales, UK became the first country in the world where fire sprinklers are compulsory in all
new homes [1]. In South Korea, high rise apartment buildings with more than 11 floors have been enforced to install fire sprinkler
systems since 2009 [2].
Depending on whether the water distribution pipes are filled with water or not, a fire sprinkler system is classified as the wet
pipe fire sprinkler system or the dry pipe fire sprinkler system [3]. The wet pipe fire sprinkler systems are installed more often
than the dry pipe fire sprinkler systems because they are more reliable, simpler, and more cost effective both in installation
and maintenance. Fig. 1 is a schematic diagram of a typical wet pipe sprinkler system. It consists of sprinkler heads connected
to a water distribution piping system and a water supply system. The water distribution pipes of the wet fire sprinkler system

⁎ Corresponding author.
E-mail address: sangheesuh0@gmail.com (S.H. Suh).

http://dx.doi.org/10.1016/j.engfailanal.2016.03.009
1350-6307/© 2016 Elsevier Ltd. All rights reserved.
112 S.H. Suh et al. / Engineering Failure Analysis 64 (2016) 111–125

Fig. 1. Schematic diagram of a typical wet pipe fire sprinkler system.

are always filled with water up to the sprinkler heads so that they are quick to react when sprinkler heads become open by the
heat of fire.
Either metallic or PVC tubes have been used for the water distribution pipes of the wet pipe fire sprinkler systems. Some
construction companies in Korea chose to use copper tubes for the wet sprinkler systems of the apartment buildings constructed
during the period of 2009 to 2013. This decision was made on the basis of better corrosion resistance and easier installation of
copper tubes compared to steel tubes. Actually, in Europe and North America, copper tubes accounted for more than 80% of all
tubes installed in water service [4]. Even though copper tubing was expected to give better performance in the wet fire sprinkler
systems as well, many apartment buildings with copper tubes installed in their fire sprinkler systems have suffered from severe
water leakage from the copper tubes starting from the year 2011 to the present time because of pitting corrosion of copper tubes.
In this study the cause of the pitting corrosion failure of copper tubes in the wet pipe sprinkler system was identified through
microscopical, structural and chemical analysis of copper tubes with pitting failure, the water and its sediments from several
apartment sprinkler systems in an apartment complex. Accelerated pitting corrosion of the copper tube in the wet fire sprinkler
system was simulated using a differential aeration cell. Finally, measures for stopping or delaying pitting corrosion of the copper
tube in the wet sprinkler system are suggested.

2. Experimental procedures

2.1. Visual inspection of the copper tubes with water leakage

Copper tubes removed from an apartment fire sprinkler system of an apartment complex due to water leakage were cut lon-
gitudinally into halves for visual inspection. Pictures of the pitting corrosion area were taken by a digital camera and the positions
of pitting corrosion were located relative to the water/air line.

2.2. Microscopical and structural analyses of corrosion pits

The morphological images and the compositions of the materials in the pitting corrosion area of the copper tubes were obtain-
ed by a scanning electron microscope (SEM) equipped with an energy dispersive X-ray spectroscope (EDS). Powder X-ray diffrac-
tometer (XRD) and micro-XRD were also used to identify the materials in the pitting corrosion area.

2.3. Analysis of the water and its sediments in the copper tubes of the sprinkler system

The chemical compositions of the water samples collected from the fire sprinkler systems of the apartment complex were ana-
lyzed by atomic absorption spectrometry (AAS), inductively coupled plasma optical emission spectrometry (ICP-OES) and ion chro-
matography (IC). Sediments from the water samples were analyzed by X-ray fluorescence spectroscopy (XRF) and powder XRD.
 S.H. Suh et al. / Engineering Failure Analysis 64 (2016) 111–125 113

3. Results and discussion

3.1. Macroscopic study of the pitting corrosion area

The copper tubes used for the fire sprinkler systems in this study are made of phosphorus deoxidized copper with the dimension
of 25 mm inner diameter and 0.89 mm thickness. Fig. 2 shows the copper tubes with water leakage. Green deposits were formed on
the outer surface of the tubes because of corrosion of copper by the leaked water. The points of water leakage are indicated by red
circles in Fig. 2. Fig. 3 shows the inside view of the tube. Because of the clear difference in darkness between the upper and lower
halves of the tube as shown in Fig. 3, it is easy to locate the position of the water/air lines. The reason why so much air was present
in the tube during operation is because air inside the tube was not vented enough when pressurized water was forced into the tube by
the water supply system. Actually, it is almost impossible to get rid of air in the tube by venting alone using the test valve shown in
Fig. 4 or other vent valves. It is because the air captured in the tube connected directly to a sprinkler head has no choice but to stay in
that tube, as schematically illustrated in Fig. 4. In some sprinkler systems of the apartment complex, it seems that the air has not been
vented at all before filling the tube with water, resulting in a large upper portion of the copper tube occupied by pressurized air.
Fig. 5 shows the location of the pits relative to the water/air line. In Fig. 5(a) all the pits are aligned in a line within 5 mm from
the water/air line while pits are within 2 mm from the water/air line in Fig. 5(b) and (c). Fig. 5(c) is the case where a relatively
small amount of air existed only at the top of the copper tube. These findings indicate that there should be a close relationship
between the pitting corrosion and the water/air line, as will be discussed in Section 3.6.

3.2. Morphology and composition of the corrosion products

Fig. 6(a) shows an overall picture of a corrosion affected area. A green colored corrosion product is covering the middle of the area. It
looks like that some of the corrosion product at the top and the bottom of this area were taken apart during operation of the sprinkler
system. The SEM image of the green colored corrosion product (Fig. 6(b)) shows bubble-shaped round particles joining together. This
corrosion product was scrubbed from the pit and analyzed by XRD. The XRD spectrum in Fig. 7 indicates that the corrosion product con-
sists of malachite (Cu2(OH)2CO3) and cuprite (Cu2O) particles. Cuprite particles seem to be attached to the bottom of the malachite layer.
Some green corrosion products were covered with a brown material as shown as a red circle in Fig. 8(a). This brown material
was peeled off from the pit for examination by SEM and EDS. This brown material consists of two differently shaped materials of
Fig. 8(b) and (c) which were identified as cuprous oxide (CuO) and malachite, respectively, by their EDS spectrums (Fig. 8(d) and
(e), respectively). It seems that a layer of malachite particles was formed under a layer of cuprous oxide particles. The malachite
particle might have grown without significant restriction, resulting in a needle like surface morphology as shown in Fig. 8(c).

3.3. Morphology and composition of the corrosion pits

The picture of a pitting corrosion area (Fig. 9(a)) was obtained after the green corrosion product shown in Fig. 6(a) was re-
moved by scrubbing. A big hemispherical pit and several smaller pits are shown at the center and at the bottom, respectively.
Panels (b) and (d) in Fig. 9, which are the SEM pictures of the big pit A at different magnifications, show that this pit consists
of many cubic particles which turned out to be cuprite (Cu2O) as described below. There are some spaces between the particles
which may play as passages for leaking water. Fig. 9(c) shows the morphology of a smaller pit B next to the big pit. This pit was
not covered by the green corrosion product when it was examined, probably because the corrosion product was taken apart some
time during operation. Fig. 9(c) shows that there is a deep channel surrounding the center area through which water might have
leaked out. The higher magnified picture (Fig. 9(e)) shows that this pit was covered by round shaped small particles. When some
areas indicated by red circles are examined carefully, it can be noticed that the round shaped particles are deposited on cubic par-
ticles. The big and small pits were examined for structural identification by micro-XRD and EDS. The micro-XRD and EDS results

Fig. 2. Copper tubes with corrosion failure. Red circles indicate the positions of water leakage.
114 S.H. Suh et al. / Engineering Failure Analysis 64 (2016) 111–125

Fig. 3. Inside view of a copper tube showing the water/air line.

Fig. 4. Schematic diagram of the wet sprinkler system showing how a pressurized air pocket is formed in the tube directly connected to a sprinkler head.

of the big pit A (Fig. 10(a) and (b)) show that the crystalline particles of the big pit A are purely cuprite (Cu2O) with Cl of 0.4 at.%
included. The fact that these particles are Cu2O is indicative of a low dissolved oxygen concentration inside the pit as discussed in
Section 3.6. The EDS analysis of the smaller pit B shows that the round shaped particles are mostly cuprous oxide (CuO)
(Fig. 10(c)). On the other hand, the micro-XRD analysis which has a deeper resolution than EDS shows that this pit actually con-
sists of CuO and Cu2O particles, which is consistent with what was observed by the SEM image (Fig. 9(e)). It seems that Cu2O

Fig. 5. Locations of pitting corrosion relative to the water/air line.

 S.H. Suh et al. / Engineering Failure Analysis 64 (2016) 111–125 115

Fig. 6. (a) Overall picture of a corrosion affected area, (b) SEM image of the green colored corrosion product.

particles were formed first under a layer of the green corrosion product and then the round shaped CuO particles were deposited
on Cu2O particles after the green corrosion product was somehow taken apart.

3.4. Morphology and composition of the area with no pitting corrosion

Fig. 11(a) shows the SEM picture of the area with no pits observed. Many small particles of about 5 μm in diameter are
deposited over the area. The EDS analysis result, Fig. 11(b), shows that these particles are mostly composed of Cu and O with
an atomic ratio of 1:1.1, meaning that these particles were cupric oxide (CuO). There are small peaks for C and Si as well. The
Si peak might have come from the sands in the water. It should be noticed that CuO particles are not densely coated on the
inner surface of the copper tube and there are many spaces between the particles, suggesting that CuO particles were not effective
enough to prevent general corrosion as well as pitting corrosion.

3.5. Analysis results of water and its sediments

A water sample (No. 107–9) collected from a fire sprinkler system which was repaired due to water leakage two years ago and
another water sample (No. 107–10) from other fire sprinkler system without any water leakage for four years since installation

Fig. 7. XRD spectrum of the green colored corrosion product consisting of malachite (Cu2(OH)2CO3) and cuprite (Cu2O).
116 S.H. Suh et al. / Engineering Failure Analysis 64 (2016) 111–125

Fig. 8. SEM/EDS analysis results of a brown material covering the corrosion pit: (a) brown material indicated by a red circle, (b) SEM image of CuO particles,
(c) SEM image of a Cu2(OH)2CO3 particle, (d) EDS spectrum of CuO particles, and (e) EDS spectrum of a Cu2(OH)2CO3 particle.

were analyzed. Both samples were a little alkaline with almost the same pH value of 8.25. This pH value is well in the so called
passivation region where the copper oxide protective layer is formed, preventing further oxidation of copper [5] (see Pourbaix
diagram of Fig. 14). Na, Cu, Zn, and Fe were analyzed by AAS while Mg, Ca and Al were analyzed by ICP/OES. Concentrations
of Cl− and SO2−4 were obtained by IC. The complete results are shown in Table 1. There is not much difference between the
two samples for all the elements analyzed. The concentrations of both Na+ and Cl− probably from NaCl were significant. It is
also possible that Cl− came from chlorination treatment of water. The concentration of SO24 − was also large. Cl− and SO24 −
ions are known to enhance corrosion of copper. Ca might have come either from the flux used for brazing copper tubes or
from the inhibitor (Ca(OH)2) added to water to decrease corrosion rate of steel tube.
The sediments from the collected water samples were analyzed by XRF and powder XRD. Referring to the XRF result (Table 2),
the sediments from the above two sources were mostly composed of Fe and Cu with small amounts of Si and Zn. Si might have
come from sands in the water. The composition of Fe in the sample No. 107–10 is almost 70% which is twice the Cu composition
while in the sample No. 107–9 the Cu composition is almost twice the Fe composition. The XRD results from both samples show
that the sediments are mostly composed of FeO(OH) and CuO particles (see Fig. 12). The height of the FeO(OH) (130) peak is
about twice that of the CuO (− 111) peak for the sample No. 107–10 while this ratio is reversed in the sample No. 107–9.
FeO(OH) might have originated from corroded steel parts such as the check valve. Since the results were so interesting, the sed-
iments from another two water samples, No. 106–15 from a sprinkler system repaired once from the water leakage and No. 106–
 S.H. Suh et al. / Engineering Failure Analysis 64 (2016) 111–125 117

Fig. 9. SEM images of a cluster of corrosion pits: (a) overall picture of the cluster with a big pit A and smaller pits, (b) low magnification image of the pit A, (c) low
magnification image of the pit B, (d) higher magnification image of the pit A, and (e) higher magnification image of the pit B.

13 from a sprinkler system without leakage were analyzed by XRF and the results are shown in Table 2. The sediments are com-
posed mainly of Cu and Fe. No. 106–15 sample’ Cu concentration was more than 80% while No. 106–13's Fe concentration was
almost 75%, which is the same trend as the No. 107–9 and 107–10 samples. From these results it can be concluded that the cop-
per tubes once repaired from water leakage must have undergone much more general as well as localized corrosion than the cop-
per tubes without leakage. In other words, they must have had more favorable condition for corrosion such as more air above the
water in the tube.
118 S.H. Suh et al. / Engineering Failure Analysis 64 (2016) 111–125

Fig. 10. EDS and micro-XRD analysis results of the big pit A and the smaller pit B: (a) EDS result of the big pit A, (b) micro-XRD result of the big pit A, (c) EDS
result of the smaller pit B, and (d) XRD result of the pit B.

3.6. Mechanism and structure of pitting corrosion in a wet fire sprinkler system

Copper pitting corrosion in the potable water distribution systems has been studied by many researchers, and it is induced by
different causes depending on the exposure condition [6]: chlorine induced pitting [7–9]; pitting induced by high-pH and low al-
kalinity water [10]; microbial induced corrosion [11]; erosion–corrosion due to higher velocities or localized turbulent flow;
pitting due to oxygen differential concentration cell corrosion [12]; excess flux remaining in the tube [15]; sulfide induced pitting;

Fig. 11. (a) SEM picture and (b) EDS analysis result of the area with no pits observed.
 S.H. Suh et al. / Engineering Failure Analysis 64 (2016) 111–125 119

Fig. 12. Powder XRD results of (a) No. 107–9 sample from the fire sprinkler system repaired two years ago due to water leakage and (b) No. 107–10 sample from
the sprinkler system with no water leakage.

Fig. 13. Schematic diagram illustrating how pressurized air induces pitting corrosion, creating the pit structure composed of a layer of malachite and cuprite (Cu2O)
crystalline particles.
120 S.H. Suh et al. / Engineering Failure Analysis 64 (2016) 111–125

Fig. 14. Pourbaix equilibrium diagram of copper-water [19], showing the regions where each of CuO, Cu2O, Cu++, and Cu is the stable phase. 0, −2, −4, −6 rep-
resents base-10 log of ion activity.

and stagnant water [13–15]. In many cases, copper pitting corrosion proceeds via coupling between oxidation of copper at small
anode sites and reduction of oxidants such as oxygen or free chlorine on large surface areas by the concentration cell corrosion
mechanism [12].
Stagnant water in the copper tubing, especially during the initial installation period, is cited as one of the important causes of
initiation of copper pitting corrosion [13–15]. Copper pitting corrosion in the wet fire sprinkler system is a special case caused by
stagnant water in the pipe where water remains stagnant for a very long time until the next service time. To our best knowledge,
there has been no report on the cause of copper pitting problem in the wet sprinkler system except a recent report claiming that
the carbon film, formed on copper tubes during extrusion followed by annealing process, causes the pinhole failure of the copper
tubes, resulting in leakage of water in the fire sprinkler systems [16]. Probably, the authors of the report did not recognize the
existence of pressurized air inside the copper tubes which must have a tremendous effect on pitting corrosion of the copper

Fig. 15. A schematic diagram showing a differential aeration cell: the differential aeration cell is composed of a large Cu cathode, a small Cu anode, and an agar gel
between the two fritted glass disks. CE, RE, WE, and DI in the diagram indicate counter electrode, reference electrode, working electrode, and deionized,
respectively.
 S.H. Suh et al. / Engineering Failure Analysis 64 (2016) 111–125 121

Fig. 16. Corrosion potential vs. immersion time of the large Cu in water purged with 1 atm O2, the small Cu anode in water deaerated with nitrogen purging, and
the small Cu anode in water deaerated and acidified with 0.1 M H2SO4.

Fig. 17. The anodic current density of the small anode Cu when electrically connected to the large cathode Cu in the differential aerated cell described in Fig.15.

tube. The two facts that there existed a significant amount of pressurized air over the water in the copper tube and all the cor-
rosion pits observed in this study were located near the water/air line mean that the pitting corrosion was caused by the water-
line corrosion. As stated by Cicek [17], “Waterline corrosion occurs due to the difference of oxygen concentrations close to the
water surface and below.” Therefore, the waterline corrosion is a type of oxygen concentration cell corrosion [18]. How waterline
corrosion of copper tubes occurred in the wet pipe sprinkler system is discussed in this section.
If air has not been vented enough out of the tube before water is forced into the tube, quite a volume of air will stay over the
water in the tube with the same air pressure as the water pressure. The water pressure in the copper tubes was from 4 to 8 atm
in the 12-story apartment of this study. The lower the level of the floor, the higher the water pressure was. Therefore, the water-
line corrosion of this study in which a given volume of pressurized air stays over the water is a little different from the usual wa-
terline corrosion in which fresh air is continuously supplied to the water/air line. Fig. 13 shows schematically how the pressurized
air in the copper tube causes pitting corrosion. Because the solubility of oxygen gas in water is proportional to the partial pressure
of oxygen over the water, a region where the dissolved oxygen concentration is very high will be formed just along the water/air
line soon after the water is filled into the tube. The areas adjacent to the water/air line will have a lower dissolved oxygen

Table 1
Analysis results of the water collected from the sprinkler systems (unit: ppm).

Sample no. Na Cu Zn Fe Mg Ca Al Cl− SO2−

4

107–9⁎ 7.46 b0.05 b0.03 b0.1 3.17 9.17 b0.1 12.61 10.70
107–10⁎⁎ 7.22 b0.05 b0.03 b0.1 3.62 14.0 b0.1 14.53 9.12
⁎ Sample from a fire sprinkler system repaired once from water leakage.
⁎⁎ Sample from a fire sprinkler system without water leakage.
122 S.H. Suh et al. / Engineering Failure Analysis 64 (2016) 111–125

Table 2
XRF analysis summary of the sediments collected from the sprinkler systems (unit: mass %).

Sample no. Fe Cu Zn Si Mg Al Ca Cr Mn Ni Pb K P S Cl

107–9⁎ 34.7 62.0 0.78 1.39 0.08 0.12 0.28 0.10 0.06 0.04 0.05 0.01 0.10 0.19 0.03
107–10⁎⁎ 69.1 25.2 1.64 2.20 0.11 0.24 0.39 0.17 0.14 0.10 0.11 0.02 0.21 0.36 0.04
106–15⁎ 16.5 81.3 0.59 0.78 0.13 0.12 0.27 0.03 0.04 0.02 0.05 0.01 0.05 0.06 0.02
106–13⁎⁎ 74.2 19.2 2.12 2.30 0.07 0.63 0.37 0.23 0.13 0.11 0.22 0.03 0.08 0.26 0.04
⁎ Sample from a fire sprinkler system repaired once from water leakage.
⁎⁎ Sample from a fire sprinkler system without water leakage.

concentration of the filled water, usually from 5 to 8 ppm. A small deposit of carbon film or iron hydroxide debris may exist on
the nearby inner surface of the copper tube submerged in water and it may block diffusion of oxygen to the space under the de-
posit or debris, forming a small oxygen deficient region there. An oxygen concentration cell can be established between the small
oxygen deficient region acting as a small anode and the region of very high dissolved oxygen concentration along the water/air
line acting as a large cathode.
Pitting corrosion of the copper tube can be explained by using the Pourbaix diagram of Cu–H2O system shown in Fig. 14 [19].
A Pourbaix diagram is basically a potential-pH diagram which is divided into several regions of stable phases such as CuO, Cu2O,
Cu2+, and Cu for Cu–H2O system. There are two dotted lines in Fig. 14: one for oxygen reduction; O2 +2H2O + 4e- → 4OH- and the
other one for hydrogen reduction; 2H+ + 2e- → H2. The potential difference between the high oxygen concentration region and
the low oxygen concentration region, ΔEO2 is given by the Nernst Eq. (1) where POhigh
2
and POlow
2
are the oxygen partial pressures
over the high and low oxygen concentration regions, respectively. There is no way to measure the oxygen concentration in the
pit. However, from the fact that Cu2O was formed underneath the Cu2(OH)2CO3 layer as shown in Figs. 9 and 10, the potential
difference is estimated to be between 0.54 and 0.74 V from the Pourbaix diagram of Cu–H2O system for PO2 = 1 atm (Pair =
5 atm) and the water pH of 8, as shown in Fig. 14.


high low
ΔΕO2 ðVÞ ¼ 0:015 log PO2 =PO2 ð1Þ

Because of this high potential difference, the oxygen reduction reaction will occur in the high oxygen concentration region acting
as a cathode while the copper in the oxygen deficient region will act as an anode and hence be dissolved into water as Cu ions, as given
by the half cell reaction Eqs. (2) and (3), respectively. In the half cell reaction of Eq. (3), Cu2+ ion is formed at the anode because it has
a higher enthalpy of hydration than its counterpart, Cu+ ion [20]. Electrons given off from the anode when copper is dissolved will be
supplied to the cathode via the copper tube, making the oxygen reduction reaction possible.
.
1 − −
O2 þ H2 O þ 2e →2OH ðat the cathodeÞ ð2Þ
2

2þ −
Cu→Cu þ 2e ðat the anodeÞ: ð3Þ

The potential of the reaction (2) is given as a function of oxygen partial pressure PO2 and pH at the cathode by Eq. (4) which
was obtained from the Nernst equation.

0:0591

− 4
Ecathode ðVÞ ¼ 0:401− log ½OH  =PO2
4 ð4Þ
¼ 1:228 þ 0:0148 logPO2 −0:0591pH

where 0.401 V is the standard equilibrium potential of the oxygen reduction reaction of Eq. (2) [21]. The potential of the reaction
(3) is given as a function of the concentration of Cu2+ by Eq. (5) which was obtained from the Nernst equation.

h i
2þ
Eanode ðVÞ ¼ 0:340  0:0296 log Cu ð5Þ

where 0.340 V is the standard equilibrium potential for the reduction reaction of Cu2+ to Cu [21], which is the reverse reaction of
Eq. (3).
The total cell reaction is given by Eq. (6) that is the summation of Eqs. (2) and (3).

.
1 2þ −
Cu þ O2 þ H2 O→Cu þ 2OH ð6Þ
2
 S.H. Suh et al. / Engineering Failure Analysis 64 (2016) 111–125 123

The total cell reaction potential is given by Eq. (7), which is actually independent of the oxygen concentration in the low ox-
ygen concentration region.

h i
2þ
ΔE ðVÞ ¼ 0:888 þ 0:0148 logPO2  0:0591 pH  0:0296 log Cu ð7Þ

For the condition of [Cu2+] = 10-2M, PO2 = 1 atm, and pH = 8, the potential for this corrosion reaction is as big as 0.47 V (see
Fig. 14). This value can be also obtained directly from the Pourbaix diagram, as shown in Fig. 14. Therefore, once an oxygen con-
centration cell is formed between the high and low oxygen concentration regions, it would be very difficult to stop dissolution of
copper in the pit.
Soon after initiation of the pitting corrosion, the dissolved oxygen concentration in the water will become uniform in the tube
by diffusion, its value depending on the water pressure. In this study, because the water pressure was about 5 atm, the dissolved
oxygen concentration will become about 40 ppm, 5 times the value at 1 atm (8 ppm). Because the dissolved oxygen concentra-
tion is now uniform throughout the tube, a much larger area can play as the cathode, continuing to dissolve copper in the pits.
Accelerated pitting corrosion of the copper tube in the wet fire sprinkler system was simulated and further discussed using a
differential aeration cell in Section 3.7.
On the basis of the morphologies and structures observed in this study, the schematic picture of the pit and its surroundings can be
drawn as shown at the top right corner of Fig. 13. There are CuO particles on top of the Cu2(OH)2CO3 malachite layer which is in con-
tact with Cu2O particles. Deep inside the pit, there will be Cu2+ ions dissolved from the copper tube. How this structure was formed
can be explained by referring to the Pourbaix diagram of Cu–H2O system [19]. Because the dissolved oxygen concentration is very low
inside the pit, Cu2O rather than CuO becomes the stable phase in the pit. Therefore, Cu2+ ions will react with H2O to precipitate Cu2O
according to Eq. (8). Cu2+ ions accumulated in the pit will exert an attractive coulombic force on Cl− ions dissolved in water outside of
the pit, and hence the Cl− ions are electromigrated into the inside of the pit, which might accelerate the dissolution of copper [22,23].
Actually, Cl− was detected when the Cu2O particles were analyzed by EDS (see Fig. 10(a)). H+ ions formed by the hydrolysis of Cu2+
ions, as shown in Eq. (8) tend to make the water inside the pit acidic, helping more dissolution of copper.

2þ þ
2Cu þ 2H2 O→Cu2 O þ 4H ð8Þ

Just outside of the pit where more OH− ions and CO2− 3 ions are available, Cu2+ ions leaking from the pit will react with these
ions to form a layer of malachite according to the following equation.

2þ 2 
2Cu þ CO3 þ 2OH →Cu2 ðOHÞ2 CO3 : ð9Þ

The malachite layer grows in size as more Cu2+ ions are leaking out of the pit, preventing oxygen at the outside of the pit
from reaching the inside of the pit and maintaining the state of low oxygen concentration inside. As the malachite layer grows
in size, the area which was initially not covered with the layer becomes covered with the malachite layer and oxygen deficient
so that new corrosion pits can be nucleated there. This can explain why a group of corrosion pits of different sizes shown in
Fig. 9 were observed under the same malachite layer.
On top of the malachite layer and on the inner surface of the copper tube where no pits were found, only CuO particles were
observed. This is because of the high dissolved oxygen concentration of the water inside the copper tube. These CuO particles
were found loosely attached to the inner surface of the copper tube so that some of them were washed away when water was
collected for analysis. That is the reason why so much CuO particles was found in the sediments of the water examined. It
seems that the layer of CuO particles was not dense enough to protect the copper surface from general corrosion as well as pitting
corrosion.

3.7. Simulation of pitting corrosion of the copper tube in the wet fire sprinkler system

Pitting corrosion of the copper tube in the wet fire sprinkler system was simulated using a differential aeration cell shown in
Fig. 15. The differential aeration cell consists of a large Cu cathode, a small Cu anode, and an agar gel (a salt bridge solution sat-
urated with KCl) between the two fritted glass disks. The large Cu cathode of 1.1304 cm2, simulating the large surface area outside
of a pit with a high dissolved oxygen concentration, is exposed to water aerated by purging with 1 atm O2 (equivalent to 5 atm
air inside the copper tube of the sprinkler system), while the small Cu anode of 0.1256 cm2, simulating the internal bottom of a
small pit where the dissolved oxygen concentration is very low, is exposed to water deaerated with N2 purging. (The area ratio of
the large Cu cathode to the small Cu anode is 9.) The large Cu cathode and the small Cu anode are physically separated from each
other but electrically connected through a zero resistance ammeter (ZRA) for electron path and the agar gel for ion path.
Fig. 16 shows the corrosion potential vs. immersion time of the large Cu exposed to water purged with 1 atm O2, the small Cu
anode exposed to water deaerated with nitrogen purging, and the small Cu anode exposed to water deaerated and acidified with
0.1 M H2SO4. The corrosion potential of the large Cu is about 80 mV noble to that of the small Cu, hence the large Cu would act as
a cathode whereas the small Cu as anode when the two electrodes are electrically connected. Hence, on the surface of cathode,
cathodic reaction or the reduction reaction of oxygen (see Eq. (2)) will occur. In contrast, on the surface of the small Cu
anode, anodic reaction of Cu (see Eq. (3)) will occur. The smaller difference (80 mV) in the corrosion potential between the
124 S.H. Suh et al. / Engineering Failure Analysis 64 (2016) 111–125

Table 3
Corrosion potential and corrosion rate (icorr) of small Cu anode in deaerated DI water and those of large Cu cathode in aerated DI water with 1 atm O2, respectively.

Small Cu anode Large Cu cathode

Deaerated DI water Aerated DI water

Ecorr (VSCE) −0.035 0.045

icorr (A/cm2) 1.82 × 10−7 2.74 × 10−7

large Cu cathode and the small Cu anode is due to the fact that the corrosion potential is not an equilibrium potential but a mixed
potential at which anodic reaction rate of Cu and the oxygen reduction rate on Cu are equal.
The corrosion potential and corrosion rate (icorr) of Cu, without electrical connection between the two electrodes, were
measured potentiodynamically using the conventional three electrode polarization cell in the same environment as that in the
differential aeration cell shown in Fig. 15, and Table 3 shows the results.
It is evident from Table 3 that the corrosion rate of large Cu in aerated DI water is faster than that of small Cu in deaerated DI
water when the two Cu electrodes are electrically disconnected. This is due primarily to the fact that corrosion rate of Cu
increased with the concentration of oxygen dissolved in water.
Fig. 17 and Table 4 show the anodic current density (corrosion rate) of the small Cu anode in deaerated water, and in
deaerated 0.1 M H2SO4 water vs. the immersion time and their average values, respectively, when the two electrodes were elec-
trically connected in the deaeration cell shown in Fig. 15. The anodic current density or the corrosion rate of small Cu anode in-
creased to 6.84 × 10−6 A/cm2 from 1.82 × 10−7 A/cm2 by electrically coupling to the large Cu cathode in the differential aeration
cell as presented in Tables 3 and 4. In fact, the area ratio of cathode to anode in an actual case of pitting corrosion would be much
greater than 9 of this simulation test because the initial site of pitting corrosion is extremely small, and hence the anodic current
density (6.84 × 10−6 A/cm2) or the corrosion rate of the small Cu anode in the deaeration cell would be increased more as the
ratio of cathode to anode increases. The anodic current of the small Cu anode would be consumed to form a localized hole or the
pit. Actually it takes about more than two years for leaking by the pitting corrosion to occur in the copper tubes of the sprinkler
system.
As the small pit initiates and grows with time, the internal water of the pit gets acidic by H+ ions produced by the hydrolysis
of Cu2+ according to Eq. (8). In water acidified with 0.1 M H2SO4 simulating the internal water of a pit at a growth stage, the
small Cu anode was 0.108 V active to the large cathode exposed to water purged with 1 atm O2, which is greater than the
value (0.08 V) measured in neutral DI water, as shown in Fig. 16. The anodic current density or the corrosion rate of the small
Cu anode in water acidified with 0.1 M H2SO4 was measured to be 9.28 × 10−6 A/cm2 as shown in Fig. 17 and Table 4 when
the two electrodes are electrically connected, indicating the growth rate of a pit formed in Cu tube increased with the acidification
of the pit internal water, thereby forming a hole eventually.

3.8. Measures to prevent pitting corrosion in the wet fire sprinkler system

In Sections 3.6 and 3.7, it was well discussed how the pressurized air pocket in the copper tube over water causes pitting cor-
rosion. In order to prevent the pitting corrosion, the air should be completely removed from the copper tubing before filling
water. Careful venting of the tubing would help reduce the occurrence of pitting corrosion. Purging the tubes with inert gas
such as nitrogen before filling the tube with water would be the other option for removing air from the tubing [24]. In this
case, even though a waterline is formed inside the tube, the gas over the waterline is inert so that no region of high oxygen con-
centration will be formed in the water.
Then, is it possible to prevent already existing corrosion pits from continuing to grow until leakage? It would be a very difficult
task because the dense layer of malachite prevents the water from being mixed between the inside and the outside of a pit so that
the state of low oxygen concentration inside the pit continues to prevail. In order to completely prevent growth of corrosion pits,
in addition to removing air from the tubing it would be necessary to reduce the dissolved oxygen concentration in the water to
such a low value that any existing oxygen concentration cell stops functioning.
The measures suggested above will be useful for preventing pitting corrosion as well as general corrosion of metallic tubes
including copper and steel tubes.

Table 4
Corrosion rates of the small Cu anode in deaerated DI water, and in 0.1 M H2SO4 dearated DI water when the two electrodes were electrically connected.

Small Cu anode Small Cu anode

in deaerated DI water in 0.1 M H2SO4 dearated DI water

Corrosion rate (A/cm2) 6.84 × 10−6 9.28 × 10−6

 S.H. Suh et al. / Engineering Failure Analysis 64 (2016) 111–125 125

4. Conclusion

Based on the analysis of the pitting corrosion failure of copper tubes in an apartment fire sprinkler system, the following
conclusions were reached:

(1) The waterline localized corrosion, a type of differential oxygen concentration cell corrosion, induced by pressurized air over
the water in the copper tube was the cause of the pitting corrosion failure of copper tubes in the wet fire sprinkler systems.
(2) A state of a very low oxygen concentration was maintained under the envelope of a dense layer of malachite formed as the
corrosion product, helping initiation of pitting corrosion. Further, both the acidification resulting from the hydrolysis of Cu
ions that occurs inside the pit and the electromigration of Cl ions into the pit would accelerate the propagation of the
pitting corrosion.
(3) CuO particles deposited on the inside of the copper tube do not seem to help prevent localized corrosion as well as general
corrosion.
(4) A measure to prevent pitting corrosion would be completely removing air from the copper tube before filling water. In
addition to this measure, the dissolved oxygen concentration in the water should be reduced to a very low value in
order to prevent already existing corrosion pits from continuing to grow.

Acknowledgments

This study was supported by the Korea Institute of Science and Technology through the KIST Institutional Program (Project no.
2E25440), and by the SH Corporation through the KIST Small Size Program (Project no. 2S01704).

References

[1] BBC News. 16 February 2011. (Retrieved 4 August 2011).

[2] Enforcement ordinance for firefighting facilities, maintenance and safety management of Korea, 2015.
[3] http://www.bafsa.org.uk/sprinkler-information/types-of-sprinkler.php.
[4] Internal Corrosion of Water Distribution System, Report of Cooperation Research, AWWA Research Foundation, USA, 1985.
[5] P.J.L. Fernandes, Type I pitting of copper tubes from a water distribution system, Eng. Fail. Anal. 5 (1998) 35–40.
[6] P. Scardina, M. Edwards, D.J. Bosch, G.V. Loganthan, S.K. Dwyer, Assessment of Non-uniform Corrosion in Copper Piping, AWWA Research Foundation, USA, 2008.
[7] B. Custalow, E. Sarver, M. Edwards, Factors influencing copper pitting induced by high pH and high chlorine, Proceedings of the 2009 AWWA Annual Conference,
2009.
[8] H. Cong, J.R. Scully, Effect of chlorine concentration and resultant cathodic half-cell kinetics on natural pitting of copper as a function of water chemistry, J.
Electrochem. Soc. 157 (2010) C200–C211.
[9] B. Marshall, M. Edwards, Copper pinhole leak development in the presence of Al(OH)3 and free chlorine, Proceedings of the AWWA Annual Conference, San
Francisco 2005, p. 16.
[10] D.A. Lytle, M.R. Schock, Pitting corrosion of copper in waters with high pH and low alkalinity, J. Am. Water Works Assoc. 100 (2008) 115–128.
[11] T.D. Burleigh, C.G. Gierke, N. Fredj, P.J. Boston, Copper tube pitting in Santa Fe municipal water caused by microbial induced corrosion, Materials 7 (2014)
4321–4334.
[12] M. Sosa, S. Patel, M. Edwards, Concentration cells and pitting corrosion of copper, Corrosion 55 (1999) 1069–1076.
[13] M. Edwards, J.F. Ferguson, S. Reiber, The pitting corrosion of copper, J. Am. Water Works Assoc. 86 (1994) 74–90.
[14] E. Sarver, M. Edwards, Inhibition of copper pitting corrosion in aggressive potable waters, Int. J. Corros. (2012), Article ID 857823.
[15] O. Richard, P.E. Lewis, A White Paper Review: History of Use and Performance of Copper Tube for Potable Water Service, https://www.wsscwater.com/water-
quality–stewardship/research/pinhole-leaks–corrosion-control/copper-pipe-white-paper.html 2016.
[16] J.-B. Lee, H. Jung, Investigation on causes of pitting corrosion in sprinkler copper tubes, Corros. Sci. Tech. 13 (2014) 6–14.
[17] Volkan Cicek, Corrosion Engineering, Scrivener Publishing Wiley, MA, 2014.
[18] W. Lynes, The oxygen concentration cell as a factor in the localized corrosion of metals, J. Electrochem. Soc. 103 (1956) 467–474.
[19] M. Pourbaix, Atlas of Electrochemical Equilibria in Aqueous Solutions, Pergamon Press, Houston, 1974.
[20] A.F. Holleman, E. Wiberf, N. Wiberg, Inorganic Chemistry, Academic Press, CA, 2001.
[21] R.C. Weast, M.J. Astle (Eds.),CRC Handbook of Chemistry and Physics, 60th Ed., D-155, Boca Raton, Fl., 1980.
[22] M. Pourbaix, Recent applications of electrode potential measurements in the thermodynamics and kinetics of corrosion of metals, Corrosion 25 (1969) 267–281.
[23] D.A. Lytle, M.N. Nadagouda, A comprehensive investigation of copper pitting corrosion in a drinking water distribution system, Corros. Sci. 52 (2010) 1927–1938.
[24] Japan's Patent, JP2003–290380, Wet sprinkler system and the prevention of corrosion of its tubing, 2003.