Corrosion of steel under the defected coating studied by localized electrochemical impedance spectroscopy

Page 1

A

(cicformp©

K

1

hmfct

EtibehE

0d

Available online at www.sciencedirect.com

Electrochimica Acta 53 (2008) 4740–4747

Corrosion of steel under the defected coating studied by localizedelectrochemical impedance spectroscopy

C. Zhong a,b, X. Tang a,c, Y.F. Cheng a,∗a Department of Mechanical and Manufacturing Engineering, University of Calgary, Calgary, Alta., T2N 1N4 Canada

b Department of Materials Science, Fudan University, 220 Handan Road, Shanghai 200433, Chinac College of Chemistry and Chemical Engineering, Ocean University of China, Qingdao, Shandong 266100, China

Received 4 December 2007; received in revised form 4 February 2008; accepted 5 February 2008Available online 12 February 2008

bstract

Corrosion of steel under the defected coating in near-neutral pH solution was investigated by localized electrochemical impedance spectroscopyLEIS) measurements. The LEIS response is dependent on the size of the defect. For small defects, e.g., less than 200 �m in diameter, localizedorrosion process and mechanism of steel, as indicated by the measured LEIS plots, change with time. The diffusion process dominates thenterfacial corrosion reaction, which is due to the block effect of the deposited corrosion product combined with the geometrical factor of a largeoating thickness/defect width ratio. In the presence of a big defect, e.g., up to 1000 �m, the LEIS responses measured at the defect are alwayseatured by a coating impedance in the high-frequency range and an interfacial corrosion reaction in the low-frequency range. The block effect

f corrosion product does not apply due to the relatively open geometry. Conventional EIS measurements on a macroscopic-coated electrodeeflect the “averaged” impedance results from both coating and defect. The information of the localized electrochemical corrosion processes andechanisms at the small defect is lost, and the coating impedance information is “averaged” out when a big defect is contained. LEIS measurement

rovides an essential technique to characterize microscopically the local electrochemical corrosion reaction of steel under the defected coating.2008 Elsevier Ltd. All rights reserved.

hemic

libc

atTotrst

eywords: Corrosion; X-65 pipeline steel; Defect; Coating; Localized electroc

. Introduction

Electrochemical impedance spectroscopy (EIS) techniqueas been used extensively to characterize the coating perfor-ance and the corrosion of steel under the coating [1–3]. Mans-

eld and co-workers [1,2,4–6] have developed typical impedanceriteria to use EIS characteristics and parameters for determina-ion of corrosion mechanism and rate of steel under coating.

Characterization of coating performance using conventionalIS measurement is associated with a major shortcoming, i.e.,

he measured impedance result is attributed to the electrochem-cal response of the whole electrode, reflecting an “averaged”ehaviour of the macroscopic electrode. As a consequence, local

lectrochemical process occurring at micro-defect such as pin-ole in the coating is “averaged” out. Therefore, analysis ofIS data is incapable of revealing the process and mechanism of

∗ Corresponding author. Tel.: +1 403 220 3693; fax: +1 403 282 8406.E-mail address: [email protected] (Y.F. Cheng).

cv

pAvi

013-4686/$ – see front matter © 2008 Elsevier Ltd. All rights reserved.oi:10.1016/j.electacta.2008.02.014

al impedance spectroscopy

ocalized electrochemical corrosion reaction of steel under coat-ng with micro-scaled defects. Without such information, it wille impossible to develop a complete understanding of localizedoating degradation and the resultant corrosion of the steel.

Localized EIS (LEIS) measurements provide a promisinglternative to investigate microscopically the coating degrada-ion and localized corrosion of steel under coating. Zou andhierry [7] used LEIS technique to study the degradation ofrganic coatings on metals. It was found that LEIS could be usedo detect the onset of coating blistering and give a measure of theate of water diffusion into the organic coating at a microscopiccale. The technique has also been applied to study the propaga-ion of pits on two different iron–chromium alloys in chloride-ontaining environments [8], demonstrating that LEIS can givealuable information on the mechanism controlling pit growth.

High performance composite coating (HPCC) is being

lanned to use for maintenance of pipeline integrity in therctic and sub-Arctic areas [9,10]. To date, there has beenery limited work to characterize electrochemically the coat-ng property for prevention of corrosion of steel, especially

Page 2

mica A

tdmisLcgmtid

2

2

mwifaaa

w

c(Mrtd5

2

PwtTew

fiwttw

C. Zhong et al. / Electrochi

he localized corrosion process normally occurring at micro-efects, such as pinholes in the coating. In this work, LEISeasurements were performed on HPCC-coated pipe steel to

nvestigate the localized corrosion process and mechanism ofteel under the defected coating in a near-neutral pH solution.EIS plots were measured either directly above the defect over aertain frequency range or by mapping the defected area at a sin-le measurement frequency. For comparison, conventional EISeasurements were also conducted on the macroscopic elec-

rode. It is anticipated that this research provides an essentialnsight into the mechanism of localized corrosion of steel underefected coating in near-neutral pH solution.

. Experimental

.1. Electrode and solution

The working electrode (WE) for electrochemical impedanceeasurements was made of a sheet of X-65 pipeline steel coatedith HPCC. The coating, with an average thickness of 1.1 mm,

s a multi-component, three-layer coating system consisting of ausion-bonded epoxy (FBE) primer, a polyethylene outer layernd a tie layer containing a chemically modified polyethylene

dhesive [11]. The coated electrode area was 37.59 cm2 with anrtificial defect of 200 or 1000 �m in diameter in the center.

The test solution used was NS4 solution, which has beenidely used to simulate the dilute electrolyte trapped between

sao1

Fig. 1. Nyquist diagrams of LEIS measured directly above the defect of 200 �m

cta 53 (2008) 4740–4747 4741

oating and the pipeline, with the chemical compositiong/l): NaHCO3 0.483, KCl 0.122, CaCl2·2H2O 0.181 and

gSO4·7H2O 0.131. The solution was made from analytic gradeeagents (Fisher Scientific) and ultra-pure water (18 M�). Allests were performed at room temperature (∼22 ◦C). To repro-uce the near-neutral pH condition, the solution was purged with% CO2 (balanced with N2 gas) prior to and through the tests.

.2. LEIS measurements

The LEIS measurements were performed on WE through aAR Model 370 Scanning Electrochemical Workstation, whichas comprised of a 370 scanning control unit, a M236A poten-

iostat, a M5210 lock-in amplifier and a video camera system.he detailed configuration of the electrode system was describedlsewhere [8,12]. The relative location of the microprobe to WEas monitored by the camera system.The scanning microprobe was operated in two modes. The

rst mode was used for LEIS measurements. The microprobeith a 10 �m tip was set directly above the defect to measure

he typical impedance response at the individual point. The dis-ance between the probe-tip and the WE surface was 50 �m,hich was adjusted and monitored through a video camera TV

upplied with the Workstation. During LEIS measurements, anc disturbance signal of 10 mV was applied to the electrode atpen-circuit potential. The measuring frequency ranged from2,500 to 0.05 Hz.

in diameter after (a) 3 h, (b) 24 h, (c) 48 h, and (d) 168 h of immersion.

Page 3

4 mica Acta 53 (2008) 4740–4747

psTlf

2

tieSfp

3

3

t2iwLccaiih

dfatiWpa

3

tspr

mpilpr

Fmo

3

Fig. 5 shows the Nyquist diagrams obtained by LEIS mea-surements at intact coating. It was seen that the LEIS responsesmeasured at the intact coating were distinctly different fromthose measured at the defects. The impedance plots were fea-

742 C. Zhong et al. / Electrochi

The second mode was used for LEIS mapping. The micro-robe was stepped over a designated area of the electrodeurface. The scanning took the form of a raster in x–y plane.he step size was controlled to obtain a plot of 16 lines × 12

ines. The ac disturbance signal was 10 mV, and the excitationrequency for impedance measurements was fixed at 50 Hz.

.3. Conventional EIS measurements

The conventional EIS measurements were conducted using aypical three-electrode system, with the whole coated-steel spec-men as WE, a saturated calomel electrode (SCE) as referencelectrode and a platinum wire as counter electrode, through aolartron 1280C electrochemical system. The EIS measurementrequency was range from 20 kHz to 0.01 Hz with an applied acerturbation of 10 mV.

. Results

.1. LEIS measurements at the defect of 200 μm

Fig. 1 shows the Nyquist diagrams measured directly abovehe defect of 200 �m at the various immersion times, i.e., 3,4, 48 and 168 h, respectively. It was seen that, after 3 h ofmmersion, the LEIS plot exhibited a capacitive loop over thehole frequency range (Fig. 1). With 24 h of immersion, theEIS plot (Fig. 1b) was characterized with two depressed semi-ircles. Upon 48 h of immersion, the impedance plot (Fig. 1c)ontained two semicircles, followed by a straight line with anpproximately 45◦ slope in the low-frequency range. When themmersion time increased to 168 h, the linear impedance dom-nated the low-frequency range and a semicircle existed in theigh-frequency range (Fig. 1(d)).

Fig. 2 shows the Bode diagrams generated from the LEISata in Fig. 1. It was seen that the impedance magnitude at lowrequency decreased remarkably from 3 to 24 h of immersion,nd then decreased continuously with the immersion time. Fur-hermore, the phase angle plots showed one maximum at 3 h ofmmersion. There were two maxima after 24 h of immersion.

ith 48 h of immersion, three maxima were observed in thehase angle plot. Finally, the phase angle showed two maximafter 168 h of immersion.

.2. LEIS measurements at the defect of 1000 μm

Fig. 3 shows the Nyquist diagrams measured directly abovehe 1000 �m-sized defect at different immersion times. It waseen that there was a similar impedance feature for all LEISlots, i.e., a smaller depressed semicircle in the high-frequencyange and a larger semicircle in the low-frequency range.

Fig. 4 shows the Bode diagrams corresponding to LEIS dataeasured in Fig. 3. It was seen that the impedance magnitude and

hase angle plots almost copied each other. The low-frequency

mpedance at 0.5 Hz was approximately 5 × 104 �, which wasower than that measured at defect of 200 �m (Fig. 2(a)). Thehase angle plots showed two maxima in the whole frequencyange.

Fi

ig. 2. Bode diagrams ((a) impedance magnitude; (b) phase angle) of LEISeasured directly above the defect of 200 �m in diameter with different times

f immersion.

.3. LEIS measurements at intact coating

ig. 3. Nyquist diagrams of LEIS measured directly above the defect of 1000 �mn diameter as a function of immersion time.

Page 4

C. Zhong et al. / Electrochimica Acta 53 (2008) 4740–4747 4743

Fmi

tf

iea(

Fm

3

ew

ig. 4. Bode diagrams ((a) impedance magnitude; (b) phase angle) of LEISeasured directly above the defect of 1000 �m in diameter as a function of

mmersion time.

ured with a capacitance loop over the whole frequency rangeor all test time periods.

Fig. 6 shows the corresponding Bode diagrams at thentact coating. It was seen that all the plots almost copied

ach other. The impedance magnitude at low frequency wasbout 107 �, much higher than those measured at the defectsFigs. 2(a) and 4(a)).

Fig. 5. Nyquist diagrams of LEIS measured at intact coating.

tsioao1

3m

oaafttr

ig. 6. Bode diagrams ((a) impedance magnitude; (b) phase angle) of LEISeasured at intact coating.

.4. LEIS mapping over the defected area

Fig. 7 shows the LEIS maps measured at 50 Hz on the coatedlectrode with a central 200 and 1000 �m defect, respectively. Itas clear that the defect area in the coated electrode generated

he lowest impedance value in the map. Furthermore, Fig. 7(b)hows a larger area with the low impedance value than Fig. 7(a),ndicating that the former electrode had a bigger defect. More-ver, the bigger defect generated a smaller impedance value ofbout 4 × 104 �, while the lowest impedance value measuredn electrode containing the smaller defect was approximately.5 × 105 �.

.5. Conventional EIS measurements on aacroscopic-coated electrode

Figs. 8 and 9 show the Nyquist and Bode diagrams measuredn a HPCC-coated steel specimen with a central 200 �m defectt different immersion times (3, 24, 48 and 168 h). There wassingle capacitive loop observed at 3 h of immersion. Upon

urther immersion, the impedance plots were characterized withwo semicircles in the high- and low-frequency ranges. Withhe increase of immersion time, the sizes of both semicircleseduced. The low-frequency impedance was about 107 �.

Page 5

4744 C. Zhong et al. / Electrochimica Acta 53 (2008) 4740–4747

Fi

sdfrw

Fi

Fig. 9. Bode diagrams ((a) impedance magnitude; (b) phase angle) of EIS mea-sured on a macroscopic specimen (HPCC-coated steel with central 200 �mdefect).

ig. 7. LEIS mapping around a defect of (a) 200 �m in diameter and (b) 1000 �mn diameter.

Figs. 10 and 11 show the Nyquist and Bode diagrams mea-ured on a HPCC-coated steel specimen with a central 1000 �mefect. All plots were featured with a smaller semicircle at high-

requency range and a bigger semicircle at the low-frequencyange. With the immersion time, the low-frequency impedance,hich was around 4 × 106 �, decreased slightly.

ig. 8. Nyquist diagrams of conventional EIS measured on a macroscopic spec-men (HPCC-coated steel with a central 200 �m defect).

Fs

4

4

t

ig. 10. Nyquist diagrams of conventional EIS measured on the macroscopicpecimen (HPCC-coated steel with central 1000 �m defect).

. Discussion

.1. Corrosion of steel under defected coating

The LEIS measurements in the present work show clearlyhat the steel under defected coating experiences corrosion pro-

Page 6

C. Zhong et al. / Electrochimica A

Fmd

cFd

2wwAc(tet

LdibbaFrs(

iov

ciFpmrawdwfttfdspm

taiiWrtittac

tLatdiirsrcspbspa

ig. 11. Bode diagrams ((a) impedance magnitude; (b) phase angle) of EISeasured on a macroscopic specimen (HPCC-coated steel with central 1000 �m

efect).

esses that are distinctly different from those at the intact coating.urthermore, the steel corrosion is dependent on the size of theefect.

When coating contains small pinholes, such as the defect of00 �m in diameter in this work, corrosion of steel at the defectill experience different processes with different mechanismsith the immersion time, as identified by LEIS plots in Fig. 1.t the early stage of immersion (3 h), the LEIS plot shows a

apacitive behaviour with a high- and low-frequency impedanceFig. 1(a)). At this stage, due to the small size of the defect,here is not sufficient time for the steel to fully expose to thelectrolyte. Therefore, the impedance response is dominated byhe dielectric property of the coating.

After 24 h of immersion, two time constants were observed inEIS plot (Fig. 1(b)). Moreover, the low-frequency impedanceecreased dramatically compared to that measured at 3 h ofmmersion, as shown in Fig. 2(a). It is the typical impedanceehaviour as that measured on steel under a degraded coatingy Mansfeld [1]. The presence of two time constants, whichre also shown by two maxima in the phase angle plot in

ig. 2(b), indicates the two interfacial reaction processes occur-ing over the measurement frequency range. The high-frequencyemicircle is associated with the pore impedance at the defectabout 1.5 × 105 �), while the low-frequency semicircle is the

ei

c

cta 53 (2008) 4740–4747 4745

mpedance response corresponding to the corrosion reactionccurring at the defect base, with a low-frequency impedancealue of about 106 �.

After 48 h of immersion, three frequency-dependent timeonstants are observed in Nyquist diagram (Fig. 1(c)), whichs also indicated by the three maxima in phase angle plot inig. 2(b). Similarly, the high-frequency semicircle is from theore impedance at the defect, with an impedance of approxi-ately 1.5 × 105 �. The two time constants in the low-frequency

ange indicate that corrosion of steel at defect is experiencingn activation–diffusion mixed-control process. The straight lineith a slope close to 45◦ in the low-frequency range is a Warburgiffusive impedance [13]. The overlap of Warburg impedanceith a semicircle in the low-frequency range shows that the dif-

usion process is contributing to the whole corrosion reaction athe base of defect. Due to the small size of defect and the rela-ively long path (about 1.1 mm) for corrosive species to diffuserom bulk solution to the defect base and, furthermore, due to theeposit of the generated corrosion product blocking the diffu-ive path, diffusion step is becoming to dominate the corrosionrocess, which is confirmed by the further immersion and LEISeasurements.With 168 h of immersion, the measured LEIS plot contains

wo time constants again, with one semicircle at high frequencynd a diffusive Warburg straight line at low frequency, as shownn Fig. 1(d). The high-frequency semicircle is from the porempedance, with a value of 1.5 × 105 �. The low-frequency

arburg behaviour shows that diffusion dominates the corrosioneaction process. The diffusion-controlled effect is primarily dueo the block effect of corrosion product generated and depositedn the small defect. It agrees with previous studies, which showedhat corrosion product formed in the pores of the coating con-rolled the corrosion reaction in a neutral solution [14], and thatdiffusive impedance behaviour was related to the deposit of

orrosion product [15].When the coating contains relatively big pinholes, such as

he defect with 1000 �m in diameter in this work, the measuredEIS at the defect is completely different from that measuredt the small defect. As shown in Fig. 3, there is a common fea-ure for all LEIS plots at the different immersion times, i.e., twoepressed semicircles over the measurement frequency range,ndicating that the local impedance response at the defected coat-ng is associated with a pore impedance and a charge-transfereaction occurring at the defect base. Due to the relatively bigize of the defect, the electrolyte could reach the steel surfaceapidly when immersed in the solution. Moreover, diffusion oforrosive species is easy through the wide path to approach theteel. Thus, the corrosion reaction is in an activation-controlledrocess. Furthermore, it is observed from Figs. 3 and 4(a) thatoth the high-frequency and low-frequency impedances increaselightly with time, which is attributed to the deposit of corrosionroduct in the defect. Due to the wide geometry of the big defectnd the relatively weak corrosiveness of NS4 solution, the gen-

rated corrosion product is not sufficient to exert the block effectn the big defect.

Generally, it is almost inevitable to introduce defects to theoating during processing and operation. The present work sug-

Page 7

4 mica

gotmOdabcsmrptow[

iriirgbd

4

sifItiocFidi

bictitteu

4

c

raasitrstttritstsotpEecotc

EdDdtitraodprci

5

dtawrcl

746 C. Zhong et al. / Electrochi

ests that the size of the defect plays a critical role in corrosionf steel under the defected coating. While the defect is rela-ively small, e.g., less than 200 �m in diameter, two factors

ust be considered in analyzing the steel corrosion process.ne is the ratio of defect width to coating thickness. A smallefect existing in a relatively thick coating normally leads tonarrow path for the corrosive species to diffuse from the

ulk solution to the steel surface. The other is the deposit oforrosion product. A small defect presents a relatively closeystem. The generation of corrosion product at the defect baseay deposit and block the pathway. As a consequence, the cor-

osion of steel at the small defect with experience differentrocesses controlled by the different mechanisms, as revealed byhe different impedance features with time. Such an evolutionf localized corrosion process of steel is normally associatedith the geometry-favoured condition where corrosion occurs

16].Where the coating contains large defects, e.g., up to 1000 �m

n diameter, an open system is created for mass transport of cor-osive species. Therefore, time-dependent corrosion behaviours not observed. Even at the early stage of immersion, thempedance response is dominated by the Faradic process occur-ing at the base of the defect. Furthermore, corrosion productenerated is relatively easy to leave the defect. Therefore, thelock effect of corrosion product will not be exerted to affect theiffusion of corrosion species.

.2. Corrosion of steel under intact coating

The LEIS plots (Figs. 5 and 6) measured at intact coatinghow a pure capacitive behaviour over the immersion periods,ndicating that HPCC provides effective protection over steelrom corrosion attack, as demonstrated by previous work [10].t is seen that all the measured EIS plots contain one singleime constant, corresponding to the coating impedance. Theres no apparent change of the impedance magnitude within 168 hf immersion. It is thus reasonable to assume that there is noorrosion occurring on the steel at the steel/coating interface.urthermore, the LEIS mappings around a defect in the coat-

ng also show that the local impedance drops remarkably at theefect, while the intact coating nearby the defect generate highmpedance values, as shown in Fig. 7.

With the significant differences of the impedance responsesetween intact coating and the defect, it is thus assumed that thempedance results measured on a macroscopic-coated electrodeontaining both intact and defected areas will not reflect exactlyhe localized corrosion process occurring at defect, as indicatedn Figs. 1, 3 and 5, which is further described by comparison withhe macroscopic EIS results. Therefore, a microscopic charac-erization of electrochemical impedance behaviour at defect isssential to reveal the localized corrosion mechanism of steelnder the coating.

.3. Comparison with macroscopic EIS results

The conventional EIS measurements on a macroscopic-oated electrode containing a defect provide distinctly different

ric

Acta 53 (2008) 4740–4747

esults from those measured by LEIS. When the coating containssmall defect of 200 �m in diameter, the EIS plot measured

fter 3 h of immersion (Fig. 8) is similar to the LEIS plot mea-ured at the defect with the same immersion period (Fig. 1(a)),.e., a capacitive behaviour with one single time constant overhe measurement frequency range. The electrolyte has not yeteached the steel surface to cause corrosion of steel at thistage. Therefore, the measured impedance is dominated byhe coating property. After 24, 48 and 168 h of immersion,he impedance diagrams are featured by two semicircles inhe high- and low-frequency ranges, representing the typicallyeported EIS feature measured on the defect-containing coat-ng system [4]. The high-frequency semicircle is attributed tohe impedance response from coating, while the low-frequencyemicircle is associated with the corrosion reaction occurring onhe steel/coating interface. Compared with the LEIS plots mea-ured at defect (Fig. 1), it is apparent that EIS measurementsn the macroscopic electrode miss some important informationhat is directly associated with the electrochemical corrosionrocesses occurring at the defect. For example, the conventionalIS plots do not reveal the mass-transport process and blockffect of corrosion product as well as the resultant change oforrosion mechanism. An “averaged” effect of EIS measuredn both coating and defect causes a completely “wiping-out” ofhe localized corrosion processes occurring at the defect in theoating.

In case of a big defect up to 1000 �m, all the measuredIS plots contain a small high-frequency semicircle (not well-efined) and a relatively big low-frequency semicircle (Fig. 9).ue to the big size of defect, the impedance response from theefect is dominant over the averaged EIS behaviour. It is seenhat the high-frequency semicircles that correspond to the coat-ng impedance are so small that it could be ignorable, comparedo the low-frequency semicircles that are from the impedanceesponse of the interfacial corrosion reaction. Actually, it is morepparent from Fig. 11 that the EIS plots approximately containne time constant only. However, the LEIS plots measured atefect represent two time constants, indicating two interfacialrocesses identified by LEIS measurements. Apparently, EISesults measured on the macroscopic electrode “average” out theoating impedance information when a big defect is containedn the coating.

. Conclusions

The LEIS responses measured at the defected coating isependent on the size of the defect. For small defects, e.g., lesshan 200 �m in diameter, localized corrosion process and mech-nism of steel, as indicated by the measured LEIS plots, changeith time. The diffusion process dominates the interfacial cor-

osion reaction, which is due to the block effect of the depositedorrosion product combined with the geometrical factor of aarge coating thickness/defect width ratio.

In the presence of a big defect, e.g., up to 1000 �m, the LEISesponses measured at the defect are always featured by a coat-ng impedance in the high-frequency range and an interfacialorrosion reaction in the low-frequency range. The block effect

Page 8

mica A

og

eccda

arvrc

A

Pc(

R

[[

[

[13] G. Walter, Corros. Sci. 26 (1986) 681.

C. Zhong et al. / Electrochi

f corrosion product does not apply due to the relatively openeometry associated with the big defect.

Conventional EIS measurements on a macroscopic-coatedlectrode reflect the “averaged” impedance results from bothoating and defect. The information of the localized electro-hemical corrosion processes and mechanisms at the smallefect is lost, and the coating impedance information is “aver-ged” out when a big defect is contained.

LEIS measurement provides an essential technique to char-cterize microscopically the local electrochemical corrosioneaction of steel under the defected coating, while the con-entional EIS measurements on a macroscopic is incapable ofevealing the localized corrosion information at defects in theoating.

cknowledgements

This work was supported by Canada Research Chairsrogram, Natural Science and Engineering Research Coun-il of Canada (NSERC) and Chinese Scholarship CouncilCSC).

[

[[

cta 53 (2008) 4740–4747 4747

eferences

[1] F. Mansfeld, J. Appl. Electrochem. 39 (1995) 187.[2] C.H. Tsai, F. Mansfeld, Corrosion 49 (1993) 726.[3] J.R. Scully, S.T. Hensley, Corrosion 50 (1994) 705.[4] F. Mansfeld, Electrochim. Acta 38 (1993) 1891.[5] C.C. Lee, F. Mansfeld, Corros. Sci. 41 (1999) 439.[6] F. Mansfeld, S. Lin, V.C. Chen, H. Shih, J. Electrochem. Soc. 135 (1986)

906.[7] F. Zou, D. Thierry, Electrochim. Acta 42 (1997) 3293.[8] I. Annergren, F. Zou, D. Thierry, Electrochim. Acta 44 (1999) 4383.[9] F. King, R. Worthingham, J. Been, G. Rubie, Proceedings of IPC 2004

International Pipeline Conference, October 4–8, 2004, Calgary, Alberta,Canada, 2004, IPC04-0082.

10] G.R. Howell, Y.F. Cheng, Prog. Org. Coat. 60 (2007) 148.11] P. Singh, J. Cox, CORROSION/2000, NACE International, Houston, TX,

2000, paper no. 00762.12] G. Baril, C. Blanc, M. Keddam, N. Pebere, J. Electrochem. Soc. 150 (2003)

B488.

14] B.J. Hepburn, L.H. Callow, J.D. Scantlebury, J. Oil Colour Chem. Assoc.67 (1984) 193.

15] L.M. Callow, J.D. Scantlebury, J. Oil Colour Chem. Assoc. 64 (1981) 175.16] C. Lin, T. Nguyen, M.E. McKnight, Prog. Org. Coat. 20 (1992) 169.