An exploratory study of the corrosion of Mg alloys during interrupted salt spray testing

Corrosion Science 51 (2009) 1277–1292

Contents lists available at ScienceDirect

Corrosion Science

journal homepage: www.elsevier .com/ locate /corsc i

An exploratory study of the corrosion of Mg alloys during interrupted saltspray testing

Ming-Chun Zhao a,b,c, Patrik Schmutz b, Samuel Brunner d, Ming Liu a,b, Guang-ling Song a, Andrej Atrens a,b,*

a Division of Materials, The University of Queensland, Brisbane, Qld 4072, Australiab Laboratory for Corrosion and Integrity, Swiss Federal Laboratories for Materials Science and Technology, EMPA, Überlandstrasse 129, CH-8600 Dubendorf, Switzerlandc School of Material Science and Engineering, Central South University, Changsha 410083, PR Chinad Laboratory for Building Technologies, Swiss Federal Laboratories for Materials Science and Technology, EMPA, Überlandstrasse 129, CH-8600 Dubendorf, Switzerland

a r t i c l e i n f o

Article history:Received 19 December 2008Accepted 9 March 2009Available online 20 March 2009

Keywords:A. MagnesiumB. PolarizationB. Weight lossB. SEMSalt spray

0010-938X/$ - see front matter � 2009 Elsevier Ltd. Adoi:10.1016/j.corsci.2009.03.014

* Corresponding author. Address: Division of Materiland, Brisbane, Qld 4072, Australia. Tel.: +61 73365374

E-mail address: [email protected] (A. Atre

a b s t r a c t

A first systematic investigation was carried out to understand the corrosion of common Mg alloys (PureMg, AZ31, AZ91, AM30, AM60, ZE41) exposed to interrupted salt spray. The corrosion rates were alsoevaluated for these alloys immersed in 3 wt.% NaCl by measuring hydrogen evolution and an attemptwas made to estimate the corrosion rate using Tafel extrapolation of the cathodic branch of the polarisa-tion curve. The corrosion of these alloys immersed in the 3 wt.% NaCl solution was controlled by the fol-lowing factors: (i) the composition of the alpha-Mg matrix, (ii) the volume fraction of second phase and(iii) the electrochemical properties of the second phase. The Mg(OH)2 surface film on Mg alloys is prob-ably formed by a precipitation reaction when the Mg2+ ion concentration at the corroding surface exceedsthe solubility limit. Improvements are suggested to the interrupted salt spray testing; the ideal test cyclewould be a salt spray of duration X min followed by a drying period of (120–X) min. Appropriate appa-ratus changes are suggested to achieve 20% RH rapidly within several minutes after the end of the saltspray and to maintain the RH at this level during the non-spray part of the cycle. The electrochemicalmeasurements of the corrosion rate, based on the ‘‘corrosion current” at the free corrosion potential,did not agree with direct measurements evaluated from the evolved hydrogen, in agreement with otherobservations for Mg.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Mg alloys have significant appeal for applications in auto trans-portation due to their low density and adequate strength to weightratio. However as corrosion is an issue [1–7], much research hasbeen carried out to document and understand the corrosion ofMg alloys in common environments like 3% NaCl solution [1–55].Our research [1–3,8–22] has indicated that the corrosion of a typ-ical multi-phase Mg alloy, in a typical solution like 3% NaCl, is con-trolled by the following three factors: (i) the composition of thealpha-Mg matrix, (ii) the composition of the other phases and(iii) the distribution of the other phases. Regarding the composi-tion of the alpha-Mg matrix, no alloying element has been identi-fied that produces a solid solution Mg alloy with a corrosion ratelower than that of Pure Mg in solutions such as 3% NaCl [1,2], incontrast to Cr in Fe–Cr alloys in which a Cr content greater than11.5% completely changes the surface film and significantly in-creases the corrosion resistance [56–58]. Thus, it is a good idea

ll rights reserved.

als, The University of Queens-8; fax: +61 733653888.

ns).

to use Pure Mg as a standard to compare the corrosion behaviourof Mg alloys. Nevertheless, the composition of the alpha-Mg matrixdoes influence the corrosion rate, most probably by having an ef-fect on the surface film, which is partially protective. The composi-tion of the alpha-Mg matrix also influences the corrosion of thealpha-Mg matrix [9] as accelerated by the micro-galvanic couplingto the second phases. Regarding the composition of the otherphases, all second phases have the tendency to cause micro-accel-eration of the corrosion of the alpha-Mg matrix [1,2,9,19], so amulti-phase alloy has a corrosion rate typically greater than thatof Pure Mg. This is another excellent reason to include Pure Mgas a standard in any research on the corrosion of Mg alloys. Secondphases associated with the impurity elements Fe, Ni, Co and Cu areparticularly effective in micro-galvanic corrosion acceleration, sothat the corrosion rates are high for Mg alloys with concentrationsof these impurity elements above their tolerance limits, as much as100� greater [1,2,9,11]. Thus the study of the other influences onthe corrosion of Mg alloys requires high purity Mg alloys; high pur-ity in the context of Mg corrosion [1,2,11] means that the Mg alloycontains less than the concentration dependent tolerance level ofthe impurity elements Fe, Ni, Co and Cu. Regarding the distributionof the other phases, the second phase can provide a barrier effect if

1278 M.-C. Zhao et al. / Corrosion Science 51 (2009) 1277–1292

it is essentially continuous and has a lower corrosion rate than thealpha-Mg matrix [9,19]; otherwise there is the tendency for thecorrosion rate to be accelerated, even for second phase particlesas small as 40 nm [22]. Intergranular stress corrosion cracking isexpected for all creep resistant Mg alloys with a continuous secondphase distribution along grain boundaries [13,15,17,18]; theapplied stress propagates stress corrosion cracking caused bymicro-galvanic corrosion of the alpha-Mg by the adjacent continu-ous second phase.

Table 1 provides typical corrosion rates [9,59] for high purityAZ91 and high purity Mg. AZ91 is one of the most popular castMg alloys with a nominal composition of 9% Al, 1%Zn, balanceMg. AZ91 has typically an alpha-Mg matrix and a significantamount of second phase particles [2,9,19]. Table 1 illustrates theinfluence of (i) metallurgical condition for the Mg alloy AZ91, (ii)the fact that salt immersion tests produce corrosion rates signifi-cantly greater than salt spray corrosion tests and (iii) that the cor-rosion rate of the alloy (AZ91) is significantly greater than that ofPure Mg. Nevertheless, the corrosion rates of high purity AZ91 islower than that of mild steel in salt spray testing [1,59]. Table 1also illustrates the rather high corrosion rates for Pure Mg andMg alloys.

Our understanding of the Mg corrosion mechanism has beendeveloped by extensive studies of Mg alloys in continuous immer-sion in salt solutions. What about other environments? Mg alloysprovide good service in air with a relative humidity less than about80% [1,2,59]. This type of ‘‘dry” service environment is typified bythe inside of the passenger capsule of a car, by the inside of an of-fice or at home. Corrosion rates are low, less than 1 lm/y in thisenvironment where corrosion could be considered to be ‘‘dry cor-rosion” rather than the much faster ‘‘aqueous corrosion” as occursby an electrochemical mechanism during immersion in salt solu-tion or during exposure to continuous salt spray. Corrosion condi-tions are more aggressive outside of the auto passengercompartment such as under body and in the auto engine compart-ment; these environments are typified as interrupted salt spray,although it must be acknowledged that there is often an issueregarding the correlation between service experience and stand-ardised continuous salt spray testing. During auto service, thereare periods of salt spray (road splash), periods of exposure to hu-mid conditions and periods of drying. What are the corrosionmechanisms under these conditions? Is there a corrosion severitycontinuum between ‘‘dry corrosion” and ‘‘aqueous corrosion”?What parameters characterise the corrosion severity continuum?Does corrosion severity scale linearly with the time fraction of‘‘aqueous corrosion”? Or, do periods of drying decrease the rateof corrosion during subsequent exposure to ‘‘aqueous corrosion”?What metallurgical influences are important for Mg alloys underthese conditions? Is there an influence of alloy composition, metal-lurgical condition, nature of surface film (that may depend on alloycomposition) and the form of corrosion (particularly micro-galvanic accelerated corrosion that is expected to occur after

Table 1Corrosion rate [mm/y] measured using 10 day exposure to ASTM B117 5% NaCl saltspray test for high purity AZ91 in various metallurgical conditions (F is as cast; T4 issolution treated: 16 h at 410 �C and water quench; T6 is solution treated and aged:16 h at 410 �C and water quench, 4 h at 215 �C) compared with the corrosion ratemeasured using 96 h exposure to 1 M NaCl solution for AZ91 in similar metallurgicalconditions (F is as cast; T4 is solution treated: 100 h at 410 �C and water quench; T6 issolution treated and aged: 100 h at 410 �C and water quench, 5 h at 200 �C). Corrosionrate for Pure Mg is given for comparison.

Test for measurement of corrosion rate F T4 T6 Pure Mg Ref

Salt spray corrosion rate [mm/y] 0.64 4 0.15 – [59]Salt immersion corrosion rate [mm/y] 16 24 6 1 [9]

breakdown of the partially protective surface film)? The presentinvestigation has been designed to start to address these issuesfor Mg alloys. This work builds on our understanding of the Mgcorrosion mechanism, which is largely based on studies of Mg al-loys in continuous salt spray and continuous immersion in saltsolutions. Better understanding of the corrosion mechanism underinterrupted salt spray conditions is needed and may push the envi-ronmental envelope for the effective use of unprotected Mg alloysin auto service.

An idealised interrupted salt spray test might consist of cyclesof salt spray interspersed with periods of constant relative humid-ity and periods of drying. Drying may consolidate the surface cor-rosion product film and thereby decrease the corrosion rate duringsubsequent exposure to wet conditions during the salt spray. Themajor variables may be (1) the fraction of time of the salt spray(if the complement is the fraction of time that results in drying)(2) the duration of the typical salt spray event, (3) the fraction oftime the sample is wet (e.g. the duration of the spray cycle plusthe time the relative humidity is above 80%) and (4) the effective-ness of the drying during the drying part of the cycle (this mightrelate to the relative humidity during the drying part of the cycle).This research aims to start to provide understanding of the testingrequirements that would provide the mechanistic understandingfor the unprotected use of Mg alloys in interrupted salt spray envi-ronments associated with auto service and to explore how the cor-rosion severity might change with details of the salt spray cycle.The corrosion behaviour of Mg alloys is compared with that of pureFe and plain carbon steel.

No alloying element produces a solid solution Mg alloy with acorrosion rate lower than that of Pure Mg in solutions such as 3%NaCl. However, there are recent indications that minor alloyingadditions like Ca may produce improved corrosion performancein interrupted salt spray [61], possibly because of a more protec-tive surface film. Consequently, this project carried out a first sys-tematic investigation to understand the corrosion of a series ofcommon Mg alloys under conditions of interrupted salt spray. Fu-ture research will be aimed to explore (i) optimization of the inter-rupted salt spray testing cycle to better elucidate the corrosionmechanism and (ii) influence of alloying additions on the corrosionMg alloys associated with interrupted salt spray conditions. Theseinvestigations aim to generate a fundamental understanding of theenvironmental limits for the unprotected use of Mg alloys in envi-ronments associated with automobile service and alternating wetand dry conditions.

Electrochemical measurements were included in the presentinvestigation to continue to explore the relationship, if any, be-tween that electrochemical measurements and direct measure-ments of the corrosion rate. Some investigations of Mg corrosion[27,53–64] continue to rely only on electrochemical measurementsdespite the known [2,10,17,22–24] issue that electrochemical mea-surements of the corrosion rate, based on the ‘‘corrosion current”at the free corrosion potential, do not agree with direct measure-ments using weight loss or evaluated from the evolved hydrogen.

2. Experimental procedure

2.1. Materials

Mg alloys specimens were cut from ingot or extrusion; Pure Mg,AZ91, AM60 and ZE41 were from cast ingot whereas AZ31 andAM30 were from extrusions. The chemical compositions are pre-sented in Table 2; these alloys are high purity [1,2,11] and so aresuitable for the study of the influences on corrosion other thanthe effect of the impurity elements. The chemical composition ofthe as-rolled pure iron was 0.30% Si max, 0.045% P max, 0.045% S

Table 2Chemical composition of the Mg alloys (wt.%, balance Mg).

Material Al Zn Mn Ce La Pr Fe Sn Pb Ni Cr Cu Zr Sr

Pure Mg 0.0066 0.008 0.0045 <0.001AZ31 2.69 0.83 0.62 0.0021 <0.002 <0.002 <0.001 <0.001 <0.002 <0.002 <0.001AZ91 8.91 0.77 0.22 0.0009 <0.002 <0.002 <0.001 <0.001 <0.002 <0.002 <0.001AM30 2.99 <0.005 0.41 0.003 <0.002 <0.002 <0.001 <0.001 <0.002 <0.002 <0.001AM60 5.84 0.027 0.29 0.0009 <0.002 <0.002 <0.001 <0.001 <0.002 <0.002 <0.001ZE41 0.004 4.59 0.02 1.05 0.48 0.12 0.006 <0.002 <0.002 <0.001 <0.001 <0.002 <0.002 <0.001

Table 4Interrupted salt spray testing.

Cycle One cycle, 2 h duration, alternativecycle with UV lamps on, alternativecycle with UV lamps off

Total test duration

A 1 min salt spray + 119 min drying/humid conditions 1 week (i.e. 168 h)B 15 min salt spray + 105 min drying/humid conditions 1 day (i.e. 24 h)

M.-C. Zhao et al. / Corrosion Science 51 (2009) 1277–1292 1279

max, balance Fe and that of the as-rolled plain carbon steel (St37)was 0.17% C max, 1.4% Mn max, 0.30% Si max, 0.045% P max,0.045% S max, balance Fe. The microstructure of the Mg alloyswas examined using scanning electron microscopy (SEM) aftermetallographic preparation by mechanical grinding successivelyto 1200 grit SiC paper, polishing successively to 0.5 lm diamond,washing with distilled water, drying with warm flowing air, etch-ing in 3% nital, washing and drying.

For the immersion tests, the specimen was encapsulated inepoxy resin with an exposed surface with dimensions20 mm � 21 mm. For the interrupted salt spray testing, the speci-men size is given in Table 3. In each case, and for both types oftests, the surface was mechanically ground to 1200 grit SiC paper,washed with distilled water, dried with warm flowing air, dried ina desiccator for 1–2 days and weighed to give the specimen weightbefore exposure, Wb. At least two specimens were used for eachtype of exposure.

2.2. Immersion testing

The corrosion behaviour for aqueous corrosion was evaluatedusing immersion tests of 12 day duration, at room temperature,in 3 wt.% NaCl aqueous solution (indicated for convenience as 3%NaCl). All solutions were made with analytical grade reagentsand distilled water. Each specimen was encapsulated in epoxy re-sin with an exposed surface with dimension 20 mm � 21 mm; thespecimen surface was prepared by mechanical grinding succes-sively to 1200 grit SiC paper and washing with distilled water;the specimen was horizontally immersed in 1000 ml of 3% NaClsolution and the corrosion rate was evaluated by measuring theevolved hydrogen, collected in a burette above the corroding sam-ple. The evolved hydrogen is a direct measure of the corrosion rate,as, in the overall magnesium corrosion reaction

MgþHþ þH2O ¼Mg2þ þ OH� þH2 ð1Þ

one molecule of hydrogen is evolved for each atom of corrodedmagnesium. The immersion test used only one working surfaceand the corrosion micro-morphology could be investigated whenusing a metallographic prepared working surface [9,10,21,19].

2.3. Interrupted salt spray testing

The corrosion behaviour was evaluated using an exposure cycleinformed by the GM test procedure GMN9319TP. The aim was asimple idealised interrupted salt spray test, which had salt sprayperiods, periods of drying conditions and periods of humid condi-tions. The modified commercial accelerated weathering tester(QUV) had eight separate, thermally-insulated chambers where

Table 3Specimen size for interrupted salt spray tests; sample thickness was 1–3 mm.

Materials Pure Mg AZ31 AZ91 AM3

Size [mm] 40 � 40 54 � 22 48 � 25 54 �

eight separate weathering experiments could be run simulta-neously; this tester is described by Brunner et al. [65]. The testerused by Brunner et al. [65] was further modified for this study sothat each sample in each chamber was subjected to essentiallythe same temperature and exposure history. The salt spray solu-tion was 3% NaCl. To exclude any possibility of contamination,fresh solution was used for each spray; i.e. the solution was notre-circulated. Specimen size is given in Table 3. A polyamide(PA6.6) screw through an 8 mm hole in the centre of each speci-men fixed the specimen to a painted Al coupon rack at a distanceof about 5 mm from the painted aluminium surface using a nylonwasher; insulating mounting was used to prevent any possibility ofa galvanic effect between the sample and its mounting; any possi-bility of crevice corrosion was avoided by the 5 mm distance be-tween the sample and the painted Al coupon rack. Each chambercontained samples of only one alloy to avoid any contaminationof the corrosion of that alloy by corrosion product run-off fromany other alloy (or from the pure iron or steel samples). Each spec-imen was placed in the chamber in a manner that ensured that theenvironment for each specimen was essentially identical duringthe test cycle, and in particular that each specimen received anessentially equal and representative quantity of the test solutionduring the spray part of the cycle. Each chamber was such thatthe spray solution should be able to freely drain from the chamberduring the whole test cycle.

The interrupted salt spray testing had 24-h operation usingtwo-hour cycles as summarised in Table 4. The temperature duringtypical cycles is presented in Fig. 1; the UV lamps were on for eachalternate two hour cycle. The choice of a 2-h cycle was mandatedby the fact that two hours was the minimum duration allowedby the commercial tester for the duration of the cycle with theUV lamps switched off. Each 2-h cycle had first a salt spray. Theduration of the spray was 1 min for Cycle A and 15 min for CycleB; the different spray durations were chosen in order to explorehow corrosion was influenced by duration of salt spray. Eachchamber was such that the spray solution would freely drain outof the chamber during and after spraying, except (it was observed)as corrosion products retained the solution. After the end of eachspray cycle, a vacuum pump sucked out the atmosphere in each

0 AM60 ZE41 Pure Fe St37

22 46 � 25 47 � 28 50 � 25 50 � 25

Fig. 1. Temperature history for two typical samples during typical portions of Cycle A (1 min salt spray + 119 min drying/humid condition). For alternative two-hour cycles,the UV lamps were either on or off.

1280 M.-C. Zhao et al. / Corrosion Science 51 (2009) 1277–1292

chamber and fresh laboratory air was allowed to flow into eachchamber. Each chamber was maintained at a temperature consid-erably above room temperature so that there would be consider-able drying of the air as the laboratory air was heated up to thetemperature of the chamber. The set temperature was 80 �C duringthe period with the UV lamps switched on and 60 �C for the alter-nate cycle with the UV lamps switched off; these were the maxi-mum allowed by the commercial tester. The expectation wasthat there would be drying during the cycle with the UV lampsswitched on whereas there would be humid conditions duringthe period with the UV lamps switched off.

2.4. Weight change evaluation

After the completion of exposure (in the solution immersion orin the interrupted salt spray), each specimen was removed fromthe environment, rinsed with distilled water, dried using flowingair, dried in a desiccator for 1–2 days and weighted to determinethe sample weight after the test with corrosion products, Wacp.An increase in sample weight is attributed to corrosion productsremaining on the specimen surface. The weight change rate, DWcp,(with corrosion products) was evaluated as:

DWcp¼fðWacp�WbÞ ½mg�g=ðspecimen area ½cm2�Þ=ðexposure time ½d�Þð2Þ

For the Mg alloy corroded specimens, the corrosion products onthe surface of the corroded samples were removed by immersionat room temperature for 5–10 min in a chromic acid cleaning solu-tion, the sample was washed with distilled water, dried with warmflowing air, dried in a desiccator for 1–2 day and weighted to deter-mine the sample weight after the test, Wam, with no corrosionproducts to allow evaluation of metal loss. The chromic acid clean-ing solution composition was 200 g/L CrO3 + 10 g/L AgNO3; previ-ous work [9,20,21] has shown that this chemical cleaningsolution causes almost no weight loss for non-corroded AZxx alloyspecimens and for the alloys used in the present work, i.e. thecleaning solution removes the corrosion products without remov-ing any significant amount of metallic Mg. Separate blank experi-ments showed that there was negligible weight change of theencapsulating epoxy resin by similar exposures to the 3% NaClsolution or to the acid cleaning solution. For the comparison pureiron and plain carbon steel corroded specimens, the corrosionproducts on the surface of the corroded samples were removedby a mild sand blast, the sample was wiped with methanol, driedwith warm flowing air, dried in a desiccator for 1–2 days andweighted to determine the sample weight after the immersion test,Wam.

The weight change rate associated with metal loss was evalu-ated as

DWm ¼fðWb�WamÞ ½mg�g=ðspecimen area ½cm2�Þ=ðexposure time ½d�Þð3Þ

2.5. Corroded surface evaluation

The corroded surface macro-morphology was recorded using adigital camera and the corroded surface was examined using scan-ning electron microscopy (SEM) after carbon coating. Energy dis-persive X-ray spectroscopy (EDS) in the SEM and X-raydiffraction (XRD) using CuKaa radiation were used to characterizecorrosion products.

2.6. Polarization curves

Potentiodynamic polarization curves were measured (immedi-ately after immersion, 24 h after immersion and 7 day after immer-sion) in 3% NaCl aqueous solution in a standard 3-electrode glasscell using a PAR-2263 potentiostat. The polarisation curves werestarted by stepping the potential several hundred mV negative tothe corrosion potential and polarising in an anodic direction at ascan rate of 0.2 mV/s. Samples for the potentiodynamic polariza-tion curves were encapsulated in epoxy resin so that a surface withthe dimension of 10 mm � 10 mm was exposed to 500 ml of solu-tion. The specimen surface was prepared by mechanical grindingsuccessively to 1200 grit SiC paper and washing with distilledwater. A platinum gauze (25 mm � 25 mm, 52 mesh) was thecounter electrode and a saturated calomel electrode (SCE) wasthe reference electrode. All potentials were referred to the SCE.

2.7. Corrosion rate evaluation

For each experiment series, the weight change rate associatedwith metal loss, DWm [mg/cm2/d], was converted to an averagepenetration rate [mm/y] using [2,9,10,23,24,60,85]

Pm ¼ 3:65DWm=q ð4Þ

where q is the metal density [g/cm3]. For Mg alloys, q is 1.74 g/cm3,and Eq. (4) becomes:

Pm-Mg ¼ 2:1DWm ð5Þ

For iron and steel, q is 7.8 g/cm3, and Eq. (4) becomes:

Pm-Fe ¼ 0:47DWm ð6Þ

M.-C. Zhao et al. / Corrosion Science 51 (2009) 1277–1292 1281

With the assumptions that (i) the corrosion of metallic mag-nesium produces corrosion product Mg(OH)2, (ii) all the corrodedmagnesium metal becomes the solid corrosion product Mg(OH)2

and (iii) all the corrosion product remains on the specimen sur-face, then the following method can be used to evaluate the cor-rosion rate from the weight change rate, DWcp. Each mol ofmetallic magnesium (24.31 g) is converted to one mol (57.31 g)of Mg(OH)2. There is a net weight increase of (57.31–24.31) g/mol. This corresponds to 24.31 g/mol weight loss of magnesiummetal. The average corrosion rate (expressed as an average pen-etration rate), can be written as

Pcp-Mg ¼ 2:1DWcpf24:31=ð57:31—24:31Þg ð7Þ

Eq. (1), the overall magnesium corrosion reaction, indicates thatone molecule of hydrogen is evolved for each atom of corrodedmagnesium. One mol (i.e. 24.31 g) of Mg metal corrodes for eachmol (i.e. 22.4 L) of hydrogen gas produced. Therefore, the hydrogenevolution rate, VH [ml/cm2/d], is related to the metallic weight lossrate, DWm [mg/cm2/d], using [2,9,10,23,24,60,85]

DWm ¼ 1:085VH ð8Þ

The corresponding penetration rate, PH-Mg, is evaluated bysubstituting Eq. (8) into Eq. (5) to give

PH-Mg ¼ 2:279VH ð9Þ

Polarization curves were used to measure the ‘‘corrosion cur-rent density”, icorr at Ecorr, by Tafel extrapolation of the cathodic

Pi-Mg ¼ 22:85icorr ð10Þ

3. Results

3.1. Interrupted salt spray cycle

Typical temperature histories for typical specimens are pre-sented in Fig. 1. Without a major modification of the apparatus,it was not possible to include humidity sensors in the chambers;these locations were used for the pipes that fed the salt solutionto the spray nozzles. Thus the conditions in a typical chamber(particularly the relative humidity and the tendency for drying)during the cycle was evaluated based on the temperature profile,as shown in Fig. 1, as well as based on considerable experiencewith the operation of the apparatus. The specimen temperature,in a typical cycle with the UV lamps switched on, increased rap-idly to a plateau value at around 55 �C indicative of a tempera-ture arrest due to the absorption of heat due to evaporation ofthe water remaining in the chamber after which there was anincrease in temperature. So the duration was about forty min-utes for the driest part of the cycle. Subsequently the specimentemperature decreased due to the salt spray at the start of thenext 2-h cycle. Thereafter, in the cycle with the UV lampsswitched off, the temperature plateau was about 45 �C, a tem-perature considerably lower than in the cycle with the UV lampsswitch on, indicating that the UV lamps provided considerableheating and that there was more effective drying during thealternate two-hour cycle during which the UV lamps wereswitched on.

The cycles were similar for Cycle B with the main differencesbeing (i) the duration of the spray was 15 min and (ii) the dryingwas considerably less effective than for Cycle A.

3.2. Microstructure

The microstructure of Pure Mg was single-phase alpha-Mg, nomicrograph has been included. Fig. 2 presents the microstructurefor the Mg alloys as documented by SEM. The microstructures ofAZ31 and AM30 were homogenous, largely single a-phase, butdid contain a small amount of fine Al/Mn particles within thea-phase matrix; the Al/Mn particle distribution was in the formof stringers elongated in the working direction as is consistent withthese samples coming from extrusions. The microstructures ofAZ91 and AM60 consisted of the a-Mg matrix and isolated coarseb-phase particles, consisted with the cast origin of these samples.The microstructure of ZE41, also consisted of the a-Mg matrixand a second phase, consistent with a cast microstructure; the sec-ond phase was located along and/or adjacent to the boundaries ofthe a-Mg dendrites in a non-continuous, somewhat-netlike distri-bution. In addition, there was a small amount of the second phasedistributed inside the a-Mg matrix as isolated small particles. Thesecond phase in ZE41 was previously identified as being consistent[21] with Mg7Zn3(RE) and Mg12RE and is different to the Mg12Al12

b-phase particles in the Mg–Al alloys.

3.3. Immersion test – hydrogen evolution

Fig. 3 presents the hydrogen evolution data for two sampleseach of each Mg alloy immersed in 3% NaCl. For Pure Mg and theMg alloys (with the exception of AZ31), there was initially an incu-bation period during which there was a small rate of hydrogen evo-lution. The incubation period was quiet long for Pure Mg, there wasessentially no incubation period for ZE41. Thereafter there was anincrease in hydrogen evolution with increasing immersion time.For most alloys, the rate of hydrogen evolution initially increasedwith increasing exposure time, which is attributed to corrosionoccurring over increasing fractions of the surface as was observedin our prior work [1,9,10,20,21]. Nevertheless, it should also benoted that the increase in hydrogen evolution could also be associ-ated with the increased actual surface area; the actual surface areaof a corroded surface is larger than the original surface area due toan increase in surface roughness. For long exposure times, the rateof hydrogen evolution became linear, attributed to steady statecorrosion and reflecting that the surface roughness had achievedsteady state. The duplicate specimens showed a close similarityof evolved hydrogen except for AZ91, for which there was amarked difference. Fig. 4 presents a comparison of the averagehydrogen evolution for each Mg alloy; each plot presents the aver-age hydrogen evolution from the two samples of each alloy. Thehighest hydrogen evolution was for ZE41 and the lowest hydrogenevolution was for Pure Mg. The hydrogen evolution can be rankedas a decreasing series: ZE41 > AM60 > AM30 > AZ91 > AZ31 > PureMg.

Table 5 presents the average hydrogen evolution data and thecorresponding average corrosion rates evaluated at 1 day, 7 dayand 12 day; in each case the corrosion rate was an average corro-sion rate evaluated using Eq. (9) and using the average hydrogenevolution rate for the relevant time period. The steady state corro-sion rate was evaluated using the final slope of the hydrogen evo-lution curve. For AZ31, the corrosion rate was essentially the samefor each time period and for steady state corrosion consistent witha nearly linear graph of evolved hydrogen. For all other magnesiumalloys, the corrosion rate increased with exposure time consistentwith the upwardly curved plots of evolved hydrogen. ZE41 showed

Fig. 2. Microstructure of each Mg alloy as revealed by SEM.

1282 M.-C. Zhao et al. / Corrosion Science 51 (2009) 1277–1292

the highest corrosion rate and Pure Mg showed the lowest corro-sion rate. The steady state corrosion rate evaluated from the hydro-gen evolution data, PH-Mg, indicated the following order for thecorrosion rate of the Mg alloys: ZE41 > AM60 > AM30 > AZ91 >AZ31 � Pure Mg. There was a similar ranking for the corrosion rateevaluated for immersion for 7 day and 12 day. The ranking was dif-ferent for the corrosion rate evaluated for 1 day immersion; thedifferences are attributed to the differences in the onset of corro-sion, see Fig. 3; it means that care is needed in the interpretationof short-term tests.

3.4. Immersion test – macroscopic corrosion morphology

Fig. 5 presents the macroscopic surface appearance of the cor-roded Mg alloys after washing with distilled water after 12 dayimmersion in 3% NaCl solution. For Pure Mg, corrosion was uniformon a macro-scale; there were no preferential sites for corrosion. Atthe end of the 12 day immersion period, the whole working surfaceof Pure Mg was homogenously covered by a layer of corrosion prod-ucts (Fig. 5a). For all other Mg alloys, the corrosion initiated at somesites on the surface, expanded over the whole surface and finally athick layer of corrosion products covered the whole working surface;

this is consistent with prior research [1,9,10,20,21]. After washingwith distilled water, the amount of visible corrosion was largestfor ZE41, Fig. 5d, and was least for AZ31, Fig. 5c, in qualitative agree-ment with the hydrogen evolution data, Figs. 3 and 4. For all Mgalloys, the corrosion manifest as wide relatively shallow areas oflocalised corrosion; as is often observed for Mg alloys, there wasno deep pitting. For AZ91, the two samples presented markedly dif-ferent hydrogen evolution volumes, Fig. 3, had significantly differentdistributions of corrosion although they showed a similar macro-scopic corrosion surface morphology, Fig. 5(e) and (f). The samplewith the lower evolved hydrogen volume had a non-corroded areaover a significant part of the sample surface, Fig. 5(f), whereas partof the other AZ91 sample had completely corroded and disappeared(i.e. for the sample with the higher hydrogen evolution), Fig. 5(e). It isalso worth noting that AM30 also showed areas of little or no corro-sion, Fig. 5(h). In contrast to AZ91, the two samples for each other Mgalloy showed a similar macroscopic corrosion surface, in qualitativeagreement with the hydrogen evolution results presented in Fig. 3;i.e. there was a close similarity of the measured data for the two spec-imens of each Mg alloy. The corrosion morphology of AM30 needsfurther elaboration; the samples had expanded from the epoxy resin,i.e. the sample surface had become higher than the surface of epoxy

0

5

10

15

AZ31

2

1

H2 e

volu

tion

volu

me,

ml/c

m2

Immersion time, h

0

10

20

30

40AZ91

2

1

H2 e

volu

tion

volu

me,

ml/c

m2

Immersion time, h

0

10

20

30

40

50

60

2

1

H2 e

volu

tion

volu

me,

ml/c

m2

Immersion time, h

AM30

0

10

20

30

40

50

60

70

21

AM60H

2 evo

lutio

n vo

lum

e, m

l/cm

2

Immersion time, h

0

50

100

150

200

2

1ZE41

H2 e

volu

tion

volu

me,

ml/c

m2

Immersion time, h

0 50 100 150 200 250 300 0 50 100 150 200 250 300

0 50 100 150 200 250 300 0 50 100 150 200 250 3000 50 100 150 200 250 300

0 50 100 150 200 250 3000

5

10

15

21

Pure Mg

H2 e

volu

tion

volu

me,

ml/c

m2

Immersion time, h

Fig. 3. Corrosion behaviour, as characterized by hydrogen evolution as a function of immersion time, for two samples of each Mg alloy.

Table 5Average hydrogen evolution and corresponding average corrosion rates for the Mgalloys immersed in 3% NaCl solution. Data are reported according to the specimen-to-specimen variability, �±10% except for AZ91 (±50%). Data precision is significantlygreater than the specimen-to-specimen variability.

Parameter Alloy 1 day 7 day 12 day Steady state

Average hydrogenevolution volume [ml]

Pure Mg 0.7 16 36AZ31 4.9 29 50AZ91 2.5 54 120AM30 8.0 110 200AM60 7.5 120 240ZE41 27 310 700

Average hydrogenevolution rate, VH

[ml/cm2/d]

Pure Mg 0.2 0.5 0.7 1.2AZ31 1.2 1.0 1.0 1.0AZ91 1 2 3 4AM30 1.9 3.8 3.9 4.1AM60 1.8 4.1 4.8 6.2ZE41 6.5 10 14 20

Average corrosionrate, PH-Mg [mm/y]

Pure Mg 0.5 1.1 1.6 2.7AZ31 2.7 2.3 2.3 2.3AZ91 1 4 6 8AM30 4.3 8.7 8.9 9.3AM60 4.1 9.3 11 14ZE41 15 24 32 46

0 50 100 150 200 250 3000

50

100

150

200

Pure MgAZ31

AZ91

AM30AM60

ZE41

Hyd

roge

n vo

lum

e, m

l/cm

2

Immersion time, h

Fig. 4. Average hydrogen evolution, as a function of immersion time, for each Mgalloy.

M.-C. Zhao et al. / Corrosion Science 51 (2009) 1277–1292 1283

resin mount, Figs. 5(h), after the 12 day corrosion immersion period.This is attributed to inner corrosion and/or corrosion product wedg-ing, in a process that might be similar to the exfoliation corrosion ofaluminium alloys; there were indications of corrosion on the backside of the sample visible through the translucent epoxy mount con-taining the specimen, indicating that some solution had penetratedto the specimen underside and corrosion product wedging couldhave been important between the epoxy mount and the specimen.

3.5. Polarization curves

Fig. 6 presents the potentiodynamic polarization curves forfreshly prepared specimens immediately after immersion in thesolution, and for specimens after 1 day and 7 days immersion in

the solution. Visual observation after the measurement of thepolarization curves revealed, in each case, that the electrode sur-face was severely corroded and had turned black and uneven.The polarization curves were not symmetrical in the anodic andcathodic branches. The rate of current increase in the anodic polar-ization branch was much greater than in the cathodic branch. Forthe polarization curves measured immediately after specimenimmersion in the solution, in the cathodic potential range, someminor localized corrosion initiated at some sites on the surface ineach case but the remainder of the surface remained bright and lit-tle changed during the polarization measurements. As reportedmany times previously [1,10], just before the corrosion potential,

Fig. 5. Representative macroscopic surface appearance of corroded Mg alloys after immersion test.

1284 M.-C. Zhao et al. / Corrosion Science 51 (2009) 1277–1292

localised corrosion initiated and expanded over the surface and thelocalized corrosion areas enlarged with increasing potential andtime. Hydrogen bubbles evolved mainly from the localized corro-sion areas. The hydrogen evolution became more intense withincreasing localized corrosion areas and with increasing potentialand time.

For the pre-corroded specimens, corrosion product covered thewhole specimen surface, and no change was visible associated withthe measurement of the polarization curve.

The non-symmetrical polarization curve between their anodicand cathodic branches in each case may be ascribed to the compli-

cated anodic polarization behaviour of Mg alloys. The surface filmaround the corrosion potential or in the anodic range is imperfect.Some areas are broken and the substrate metal is exposed to thesolution directly. There might be two reasons for the complicatednature of the anodic polarization curves. One reason is the simul-taneous combination of both anodic dissolution and cathodichydrogen evolution in the anodic region. Another reason is theoccurrence of localized corrosion, which could make the anodicprocess unstable.

The corrosion rate was evaluated from the polarization curvesby Tafel extrapolation using the cathodic branch, which was linear

-6 -5 -4 -3 -2

-1.8

-1.7

-1.6

-1.5

-1.4

-1.3

-1.2

Pote

ntia

l, V SC

E

Log (current density) or log (I), A cm-2

Pure Mg 0 min 1 day 7 days

-6 -5 -4 -3 -2

-1.8

-1.7

-1.6

-1.5

-1.4

-1.3

-1.2

Pote

ntia

l, V SC

E

Log (current density) or log (I), A cm-2

AZ31 0 min 1 day 7 days

-6 -5 -4 -3 -2

-1.8

-1.7

-1.6

-1.5

-1.4

-1.3

-1.2

Pote

ntia

l, V SC

E

Log (current density) or log (I), A cm-2

AZ91 0 min 1 day 7 days

-6 -5 -4 -3 -2

-1.8

-1.7

-1.6

-1.5

-1.4

-1.3

-1.2

Pote

ntia

l, V SC

E

Log (current density) or log (I), A cm-2

AM30 0 min 1 day 7 days

-6 -5 -4 -3 -2

-1.8

-1.7

-1.6

-1.5

-1.4

-1.3

-1.2

Pote

ntia

l, V SC

E

Log (current density) or log (I), A cm-2

AM60 0 min 1 day 7 days

-6 -5 -4 -3 -2

-1.8

-1.7

-1.6

-1.5

-1.4

-1.3

-1.2

Pote

ntia

l, V SC

E

Log (current density) or log (I), A cm-2

ZE41 0 min 1 day 7 days

Fig. 6. Polarization curves in 3% NaCl for the Mg alloys.

M.-C. Zhao et al. / Corrosion Science 51 (2009) 1277–1292 1285

in nearly all cases (a slope similar to the other plots was used in thefew cases of curved Tafel plots). The anodic branch was not used.Table 6 presents the corrosion rate evaluated from the icorr data,in comparison to the corrosion rate for the immersion tests andthe interrupted salt spray tests using Cycle A and Cycle B. Therewas considerable change with prior exposure time for the corro-sion rate evaluated from the icorr data, consistent with the estab-lishment of corrosion steady state. The data appear anomalousfor the corrosion rate for AZ31 after 7 day immersion, howeveran examination of the polarization curve, Fig. 6, indicates that

the low value is indeed measured from the polarization curve. Ifthe data for AZ31 are disregarded, then the corrosion rates forthe Mg alloys can be ranked as ZE41 > AM60 > AM30 > AZ91 > PureMg; this ranking is similar to that for the immersion tests.

3.6. Interrupted salt spray – Cycle A

Table 7 presents corrosion rates evaluated from the averageweight change rate for exposure to Cycle A: 1 min spray + 115 mindrying/humid conditions. For all the Mg alloys, the corrosion rate

Table 6Corrosion rate [mm/y] for Mg alloys estimated from the polarization curves and Tafelextrapolation to estimate icorr, (designated as PC: 0 min, PC: 1 d, PC:7 d) comparedwith the corrosion rate measured from the immersion tests (Table 5), Cycle A (Table7) and Cycle B (Table 8).

Condition Pure Mg AZ31 AZ91 AM30 AM60 ZE41

Cycle A (1 min spray) 7 d 0.6 0.8 2.3 4.0 7.4 47Cycle B (15 min spray) 1 d 1.5 2.5 0.6 2.7 1.5 2.7Immersion, 1 d 0.5 2.7 1.4 4.3 4.1 14.8Immersion, 7 d 1.1 2.3 4.1 8.7 9.3 23.7Immersion, 12 d 1.6 2.3 5.7 8.9 10.9 31.5Immersion, steady state 2.7 2.3 8.0 9.3 14.1 46.3PC: 0 min 0.9 0.7 0.2 0.5 0.2 1.1PC: 1 d 1.8 0.7 0.2 1.2 0.9 2.3PC: 7 d 0.9 0.09 1.4 2.3 7.3 9.1PH-MgðSSÞ=PiM gðSSÞ 3.0 26 5.7 4.0 1.9 5.1

1286 M.-C. Zhao et al. / Corrosion Science 51 (2009) 1277–1292

evaluated from the metallic weight loss, Pm-Mg, was always larger,and typically significantly larger, than the corrosion rate, Pcp-Mg,evaluated from the weight change rate, DWcp, (with surface corro-sion products). This indicates that a significant weight of the corro-sion products was lost from the surface and that the parameter, Pcp-

Mg, has little physical significance. Furthermore there are implica-tions on the mechanism of formation of the film on the surfaceof Mg alloys, This is discussed in Section 4.3.

The corrosion rate evaluated from the metallic weight loss after7 day exposure, Pm-Mg, indicated the following order for the corrosionrate of the Mg alloys: ZE41 > AM60 > AM30 > AZ91 > AZ31 > PureMg. This is the same ranking as produced by the immersion tests insteady state. Pure Mg, AZ31 and AZ91 had corrosion rates that werecomparable to or less than the corrosion rates of Fe and St37.

Fig. 7 presents representative views of the macroscopic surfaceappearance of the corroded Mg alloys after removal of the corro-sion products using the chromic acid cleaning solution. Theappearance is consistent with the ranking of the corrosion behav-iour with the amount of visible corrosion producing the same rank-ing of the alloys: ZE41 > AM60 > AM30 > AZ91 > AZ31 > Pure Mg.

Fig. 8 presents typical SEM views for the Mg alloys after 1 dayexposure to Cycle A. Examination was carried out after 1 day toexamine the early stages of corrosion. In all cases the appearanceis interpreted as revealing a surface covered with a cracked layerof corrosion products on top of which there were deposits of fur-ther corrosion products as well as sometimes crystals of sodiumchloride. Extensive EDX measurements were carried out. In areasof the deposits of corrosion products, the analyses typically con-tained Mg, O and Cl, with the Mg:O ratio consistent with Mg(OH)2;for AZ31 there was also some Al in a few cases and Zn in one case;for AZ91, AM30, AM60 and ZE41 there was also some Al in a fewcases but no cases with Zn, Mn or RE. Between the deposits, theEDX analyses of the corrosion product film typically containedMg, O and Cl and gave Mg:O ratios such as 7:3 or 8:2; these ratiosare interpreted as indicating a relatively thin corrosion productfilm, �1–2 lm in thickness so that there was considerable signalfrom the underlying metal; in these cases there was also signalsfrom the alloying elements: Al for AZ31, AM30 and AM60; Al andZn for AZ91; and Zn, Ce and La for ZE41.

Table 7Weight change rate data and corrosion rate for Mg alloys subjected to interrupted salt sp

Pure Mg AZ31 AZ91

Exposure time [d] 1 7 1 7 1DWcp [mg/cm2/d] 0.6 0.3 0.9 0.2 0.4Corrosion rate, Pcp-Mg [mm/y] 0.9 0.5 1.4 0.3 0.6DWm [mg/cm2/d] – 0.3 – 0.4 –Corrosion rate, Pm-Mg or Pm-Fe [mm/y] – 0.6 – 0.8 –

3.7. Interrupted salt spray – Cycle B

Table 8 presents corrosion rates evaluated from the averageweight change rate for exposure for 1 day to Cycle B: 15 minspray + 105 min drying/humid conditions. As for Cycle A, for all theMg alloys, the corrosion rate evaluated from the metallic weight loss,Pm-Mg, was always larger than the corrosion rate, Pcp-Mg, evaluatedfrom the weight change rate, DWcp, (which included the surface cor-rosion products). This indicates that a significant amount of the cor-rosion products were lost from the surface and that the parameter,Pcp-Mg, has little physical significance.

The corrosion rate evaluated from the metallic weight loss, Pm-Mg,were comparable in magnitude to those produced by Cycle A. Cycle Bproduced the following ranking for the corrosion rate of the Mg al-loys: ZE41 � AM30 � AZ31 > AM60 � Pure Mg > AZ91. This rankingis very different to that produced by Cycle A which is essentially thesame ranking as produced by the immersion tests. The differentranking in Cycle B may be attributed to differences in the onset ofcorrosion; this is particularly of significance for Cycle B becausethe 1 day test duration was relatively short, too short for the estab-lishment of steady state corrosion conditions.

Fig. 9 presents the macroscopic surface appearance of the repre-sentative corroded Mg alloys after chromic acid cleaning. Theappearance of these specimens is consistent with a lower amountof corrosion and consistent with their lower rate of corrosion.

4. Discussion

4.1. Comparison with literature

The data of Table 1 provide a good benchmark against which tocompare the present results. Table 1 indicates a salt immersioncorrosion rate of 1 mm/y for Pure Mg in 1 M NaCl measured usingimmersion tests of 96 h duration. Table 5 and Fig. 3, of the presentinvestigation, indicate that the corrosion rate of Pure Mg was0.5 mm/y for an exposure of 1 day increasing to 1.0 mm/y for a7 day exposure in good agreement with the prior literature datapresented in Table 1. In the present study, the corrosion of thetwo AZ91 specimens was very different so that the average ratepresented in Table 5 was lower than that reported in Table 1; how-ever the higher corrosion rate of the two specimens in the presentinvestigation is in good agreement with the data of Table 1. Thisgives confidence in the results of the present investigation.

The salt spray corrosion rate for as-cast AZ91 from the inter-rupted salt spray test Cycle A (2.3 mm/y, Table 6 and 7) were ob-tained in tests of 7 day duration, would be expected to be closeto steady state corrosion and thus can be compared with the10 day salt spray test corrosion rate of Table 1 for as-cast AZ91(0.64 mm/y). The corrosion rate of 2.3 mm/y measured in the inter-rupted salt spray test is significantly larger than the rate of0.64 mm/y measured in the continuous salt spray test. This differ-ence is attributed to the differences in corrosion produced by thedifferent testing conditions. The corrosion rate from the continu-ous salt spray test 0.64 mm/y is significantly lower than the rateof 16 mm/y from the continuous salt immersion test. The lower

ray testing using Cycle A (1 min spray).

AM30 AM60 ZE41 Fe St37

7 1 7 1 7 1 7 7 70.6 0.7 0.2 2.0 1.5 2.1 15.4 1.5 0.70.9 1.1 0.3 3.1 2.3 3.2 23.8 – –1.1 – 1.9 – 3.5 – 22.4 3.7 4.22.3 – 4.0 – 7.4 – 47.0 1.7 2.0

Fig. 7. Representative macroscopic surface appearance of the corroded Mg alloys after 7 day exposure to Cycle A (1 min spray).

M.-C. Zhao et al. / Corrosion Science 51 (2009) 1277–1292 1287

corrosion rate in the continuous salt spray test is attributed to thefact that in the salt spray test, there is a thin solution layer on thespecimen surface, that this thin surface layer can be considered tobe largely stagnant, and consequently the local pH increases fairlyrapidly as the Mg alloy corrodes, with the precipitation of Mg(OH)2

at the alloy surface increasing the protective nature of the surfacefilm and causing a relatively lower corrosion rate compared withthe continuous salt immersion test. In contrast, the weight changedata of Table 7 indicate that the interrupted salt spray test is char-acterized by the loss of much of the corrosion product, Mg(OH)2,formed on the specimen surface. The implication is that there werefairly aggressive salt spray conditions during the interrupted saltspray test, Cycle A, so that the salt spray solution washed awaymuch of the corrosion products and the surface film was not par-ticularly protective. Furthermore, the relatively high corrosion rateproduces by Cycle A is not consistent with corrosion occurring onlyduring the 1 min salt spray and the corrosion stopping during theremainder of the cycle (119 min drying/humid conditions). It is

much more likely that significant corrosion continued during the‘‘119 min drying/humid conditions” part of Cycle A. The implica-tion is that the interrupted salt spray test needs some improve-ments to be able to study the influence of drying during the‘‘non-salt-spray” part of the cycle. The suggested changes are out-lined in a following section.

Section 3.7 presented the results for the specimens exposedfor 1 day to interrupted salt spray – Cycle B. The results indi-cated that the testing was characterised by the onset of corro-sion rather than being characterised by steady state corrosionconditions. Thus it is not sensible to compare the results of CycleB with the steady state corrosion tests for the continuous saltspray tests of Table 1.

4.2. Comparison with steel

Despite the limitations of the Cycle A interrupted salt spray test,Table 7 shows that the corrosion rate of Pure Mg, AZ31 and AZ91

Fig. 8. SEM examination of the surface of the Mg alloys after 1 day exposure to Cycle A (1 min spray).

Table 8Weight change rate data and corrosion rate for Mg alloys subjected to interrupted salt spray testing using Cycle B (15 min spray).

Parameter Pure Mg AZ31 AZ91 AM30 AM60 ZE41 Pure Fe St37

DWcp [mg/cm2/d] 0.2 0.3 0.04 0.2 0.3 0.8 1.4 1.4Corrosion rate, Pcp-Mg [mm/y] 0.3 0.5 0.06 0.3 0.5 1.2 – –DWm [mg/cm2/d] 0.7 1.2 0.3 1.3 0.7 1.3 5.0 6.7Corrosion rate, Pm-Mg, Pm-Fe [mm/y] 1.5 2.5 0.6 2.7 1.5 2.7 2.4 3.2

1288 M.-C. Zhao et al. / Corrosion Science 51 (2009) 1277–1292

were lower or comparable to the corrosion rate of pure iron andmild steel. This is consistent with the known good corrosion per-formance of high purity Mg alloys in salt spray tests and similaratmospheric exposures in comparison with steels [1,59].

4.3. Surface film formation

The data from Cycle A, Table 7 and Cycle B, Table 8 indicate that,for all the Mg alloys, the corrosion rate evaluated from the metallicweight loss, Pm-Mg, was always larger, and typically significantlylarger, than the corrosion rate, Pcp-Mg, evaluated from the weightchange rate, DWcp, (with surface corrosion products). This indi-cates that a significant weight of the corrosion products was lostfrom the surface and that the parameter, Pcp-Mg, has little physicalsignificance. Furthermore there are implications on the mechanismof formation of the film on the surface of Mg alloys. The implicationis that the Mg(OH)2 surface film forms by a precipitation reactionwhen the Mg2+ ion concentration at the corroding surface exceedsthe solubility limit. This means that the film formation during thecorrosion of Mg alloys is somewhat similar to the formation of thepatina on copper alloys exposed to atmospheric corrosion [66,67],

the formation of transpassive film on stainless steels [68] or theformation of film on lead during anodic oxidation in concentratedsulphuric acid [69–71], rather than the growth of the surface filmas occurs for stainless alloys [56–58,72–79].

4.4. Interrupted salt spray testing

Improvements are needed to the interrupted salt spray testingbased on the evaluation of the results of Cycle A (Table 7) in com-parison with the continuous salt spray testing results (Table 1) asdiscussed above. An ideal interrupted salt spray test cycle mightbe a salt spray of duration X min followed by a drying period of(120–X) min. This would allow study of the influence of the dryingperiod. The operation of the interrupted salt spray apparatus hasbeen examined in detail to determine how to effect the desired cy-cle. It is considered that the drying can effected by the followingmodifications (i) a much reduced flow to the salt spray nozzlesso that the spray volume is much less (the lower flow rate woulddecrease significantly the washing action of the salt solution flow-ing over the specimen surface), (ii) change of the salt spray from afairly aggressive spray to a salt fog, (iii) change the apparatus pro-

Fig. 9. Representative macroscopic surface appearance of the corroded Mg alloys after 1 day exposure to Cycle B (15 min spray).

M.-C. Zhao et al. / Corrosion Science 51 (2009) 1277–1292 1289

gram so that the non-UV cycle is eliminated (the UV-on cycle hadmuch more effective drying in the present study) and (iv) increasethe amount of air exchange after the salt spray to effectively re-move all the remaining salt solution and all the remaining humidair associated with the salt spray part of the cycle and replace withdry laboratory air that would be further dried by heating to the settemperature of 80 �C. These changes would be expected to ensurethat there was drying during the non-spray part of the cycle, andthis is expected to be verified by the introduction of sensors for rel-ative humidity. The target is to achieve 20% relative humidity (RH)rapidly within several minutes after the end of the salt spray and tomaintain the RH at this level during the non-spray part of the cycle.

4.5. Influence of alloying and microstructure

Each microstructure of the Mg alloys studied in present investi-gation can be characterised as consisting of essentially two phases:an alpha-Mg matrix plus a second phase. The second phase was ineach case distributed throughout the microstructure as isolated

particles and did not form a continuous network. Thus, the priorunderstanding of Mg corrosion [1,2,9–11] leads to the expectationthat the factors controlling the corrosion behaviour are (i) the com-position of the alpha-Mg matrix, (ii) the volume fraction of secondphase and (iii) the electrochemical properties of the second phase.The corrosion of the Mg alloys in this study is, in fact, consistentwith the corrosion being controlled by these three factors as dis-cussed below.

Firstly, consider the sequence Pure Mg, AZ31 and AZ91. The al-pha-Mg phase of AZ31 has a similar composition to that of AZ91, sothe main difference between these two alloys is an increasing vol-ume fraction of the second-phase (some Al/Mn phase in AZ31 anda significant amount of beta-phase in AZ91). In nearly all casesthere is an increasing corrosion rate in the same sequence (in boththe immersion tests, Table 5, and the interrupted salt spray Cycle A,Table 6); this increasing corrosion rate is attributed to a greater de-gree of micro-galvanic corrosion because of the greater volumefraction of the second-phase. The one exception data is that thecorrosion rate of AZ31 (2.3 mm/y) appears slightly less than the

1290 M.-C. Zhao et al. / Corrosion Science 51 (2009) 1277–1292

corrosion rate of Pure Mg in steady state (2.7 mm/y) in the immer-sion test, Table 5; this difference is not statistically significant.

Similarly, there is an increase in corrosion rate in the sequencePure Mg, AM30 and AM60, which also correlates with the increas-ing amount of second-phase (attributed to causing increasingamounts of micro-galvanic acceleration of the corrosion of the al-pha-Mg matrix). The higher corrosion rates for the AMxx alloyscompared with the AZxx alloys could be due to (i) the AMxx alloyshaving a beta-phase that is more effective as a cathode for thehydrogen evolution reaction or (ii) the AZxx having a surface filmmore resistant to micro-galvanic corrosion compared with theAMxx alloys.

Similarly the higher corrosion rate of ZE41 compared with PureMg is attributed to the increased volume fraction of second phaseand the concomitant increased micro-galvanic corrosion. The high-er corrosion rates for the ZE41 compared with AZ91 and AM60could be due to (i) the effectiveness as a cathode of the secondphase in ZE41 or (ii) ZE41 having a surface film very susceptibleto micro-galvanic corrosion.

Section 3.7 presented the results for the specimens exposed for1 day to interrupted salt spray – Cycle B. The results indicated thatthe testing was characterised by the onset of corrosion rather thanbeing characterised by steady state corrosion conditions. These re-sults indicate that AZ91, AM60 and ZE41 were relatively resistantto the initiation of corrosion because their corrosion rates wheremuch less than would be expected by comparison of the compar-ative performance in both the Cycle A testing and the immersiontesting. In contrast, Fig. 3 indicates that Alloys AZ91, AM60 andZE41 were not particularly resistant to the onset of corrosion inthe immersion tests as they did not have a noticeably longer time,at short immersion times, to the initiation of more rapid corrosion.Thus the apparent resistance to corrosion of AZ91, AM60 and ZE41in the Cycle B results could be a peculiarity of that particular cycle.

4.6. Polarization curves

There were significant differences between the polarisationcurves measured for freshly prepared samples and those measuredafter the establishment of steady state corrosion conditions, Fig. 6.This indicates that there are significant differences in the details ofthe various electrochemical reactions under these two conditions.For the freshly prepared samples, the cathodic reaction of hydro-gen evolution occurs on the air formed film on the surface of thefreshly prepared surface. As shown by a number of studies[24,80–83], the corrosion potential in chloride containing solutionsrelates to the breakdown of the air formed film and the initiation ofpitting corrosion. For the sample that had attained steady statecorrosion conditions, the cathodic reaction, also assumed to behydrogen evolution, occurs on the corroded surface, possibly onthe surface of corrosion products, possibly Mg(OH)2. Clearly forPure Mg, AZ91, AM60 and ZE41, hydrogen evolution was much fas-ter on the surface after steady state corrosion has been attained,whereas there was little difference for AZ31 and AM30.

For Pure Mg, AZ91 and ZE41, the steady state corroding sur-face had a free corrosion potential more positive than that of thefreshly prepared surface and moreover the steady state corrod-ing surface had a higher rate of hydrogen evolution. This meansthat, if a corroding area is adjacent to a non-corroded area, therewill be a galvanic cell causing the galvanic acceleration of thecorrosion rate of the non-corroded area. Thus, once corrosionstarts, there is an electrochemical driving force for the spreadof the corrosion across the surface. This is indeed what is ob-served experimentally to happen. The galvanic acceleration ofthe corrosion of the non-corroding areas (i.e. the spread of cor-rosion across the non-corroding areas), is balanced by the gal-vanic protection of the corroded areas, so that the corrosion

tends to be rather shallow in the corroded areas. This is alsoconsistent with our observations [9,10]. This spread of corrosionacross the surface for ZE41 in the present study was similar tothe observation for MEZ by Song [24]. In contrast, there was lessscope for such galvanic interaction for AZ31, AM30 and AM60.

4.7. Corrosion rate from polarization curves

The electrochemical measurements were included to explorethe relationship between the electrochemical measurements ofthe corrosion rate, based on the ‘‘corrosion current” at the free cor-rosion potential, and direct measurements using weight loss orevaluated from the evolved hydrogen, because of the known[2,10,17,22–24] issue that electrochemical measurements of thecorrosion rate, based on the ‘‘corrosion current” at the free corro-sion potential, do not agree with direct measurements usingweight loss or evaluated from the evolved hydrogen. Table 6showed that the corrosion rate evaluated from the ‘‘corrosion cur-rent” at the free corrosion potential did not agree with direct mea-surements using weight loss or evaluated from the evolvedhydrogen.

Of most concern may be that there did not appear to be anyrelation between the corrosion rate at steady state measured inthe immersion tests, PH-MgðSSÞ, and the corrosion rate evaluatedfrom the polarisation curves in steady state at 7 day, PiM gðSSÞ. Thereis substantial evidence [2,9,80] that PH-MgðSSÞ is a valid measure ofthe corrosion rate for Mg alloys, equivalent to that of weight lossrate.

The ratio PH-MgðSSÞ/PiM gðSSÞ varied seemingly randomly between26 and 1.9 (the values are given in Table 6). The ratio PH/Piss wouldbe expected to be a constant if the only factor was the negative dif-ference effect and if k was a constant, where k is the fraction of theuni-positive Mg ion, Mg+, that undergoes chemical reaction withwater to liberate hydrogen. However, it is worth rememberingthe work of Petty et al. [84] indicated that k need not be a constant.In fact, due to the ‘‘anodic hydrogen evolution” phenomenon, thecorrosion rate of a magnesium alloy estimated from its polariza-tion curve is in theory not reliable [24]. The comparison of corro-sion rates obtained by hydrogen evolution and polarization curvemeasurements in this study further confirmed the ‘‘anodic hydro-gen evolution” effect [24].

5. Conclusions

1. The corrosion behaviour of the Mg alloys (Pure Mg, AZ31, AZ91,AM30, AM60 and ZE41) immersed in the 3% NaCl solution wascontrolled by the following factors (consistent with currentunderstanding of the corrosion of Mg alloys): (i) the composi-tion of the alpha-Mg matrix, (ii) the volume fraction of secondphase and (iii) the electrochemical properties of the secondphase.

2. The Mg(OH)2 surface film on Mg alloys is probably formed by aprecipitation reaction when the Mg2+ ion concentration at thecorroding surface exceeds the solubility limit.

3. Improvements are needed to the interrupted salt spray testingbased on the evaluation of the present results. The ideal inter-rupted salt spray test cycle might be a salt spray of durationX min followed by a drying period of (120–X) min. Apparatuschanges are suggested that would be expected to ensure thatthere was drying during the non-spray part of the cycle. Thetarget is to achieve 20% relative humidity (RH) rapidly withinseveral minutes after the end of the salt spray and to main-tain the RH at this level during the non-spray part of thecycle.

M.-C. Zhao et al. / Corrosion Science 51 (2009) 1277–1292 1291

4. The electrochemical measurements of the corrosion rate,based on the ‘‘corrosion current” at the free corrosion poten-tial, did not agree with direct measurements evaluated fromthe evolved hydrogen, in agreement with prior observationsfor Mg.

Acknowledgements

This work was supported by the Australian Research Centre,Centre of Excellence Design of Light Alloys. Zhao, Liu and Atrensthank EMPA for their support that allowed them to spend consid-erable periods in the EMPA laboratory. EMPA is the Swiss FederalLaboratories for Materials Science and Technology.

References

[1] G. Song, A. Atrens, Corrosion mechanisms of magnesium alloys, AdvancedEngineering Materials 1 (1999) 11.

[2] G. Song, A. Atrens, Understanding magnesium corrosion mechanism: aframework for improved alloy performance, Advanced Engineering Materials5 (2003) 837.

[3] G. Song, A. Atrens, Recent insights into the mechanism of magnesium corrosionand research suggestions, Advanced Engineering Materials 9 (2007)177–183.

[4] G.L. Makar, J. Kruger, Corrosion of magnesium, International Materials Reviews38 (1993) 138.

[5] J.D. Hanawalt, C.E. Nelson, J.A. Peloubet, Corrosion studies of magnesium andits alloys, Trans. AIME 147 (1942) 273.

[6] K.N. Reichek, K.L. Clark, J.E. Hillis, Controlling the salt water corrosionperformance of magnesium AZ91 alloy, SAE Technical Paper 850417 (1985).

[7] W.E. Mercer II, J.E. Hillis, The critical contaminant limits and salt watercorrosion performance of magnesium AE42 alloy, SAE Technical Paper 920073(1992).

[8] A. Atrens, W. Dietzel, The Negative Difference Effect and Unipositive Mg,Advanced Engineering Materials 9 (2007) 292–297.

[9] M.C. Zhao, M. Liu, G. Song, A. Atrens, Influence of the beta-phase morphologyon the corrosion of the Mg alloy AZ91, Corrosion Science 50 (2008) 1939.

[10] M.C. Zhao, M. Liu, G. Song, A. Atrens, Influence of pH and chloride ionconcentration on the corrosion of Mg alloy ZE41, Corrosion Science 50 (2008)3168.

[11] M. Liu, P.J. Uggowitzer, A.V. Nagasekhar, P. Schmutz, M. Easton, G. Song, A.Atrens, Calculated phase diagrams and the corrosion of die-cast Mg-Al alloys,Corrosion Science 51 (2009) 602.

[12] M. Liu, P.J. Uggowitzer, P. Schmutz, A. Atrens, Calculated phase diagrams, Irontolerance limit and corrosion of Mg–Al alloys, JOM 60 (12) (2008)39–44.

[13] M. Liu, D. Qiu, M.C. Zhao, G. Song, A. Atrens, The effect of crystallographicorientation on the active corrosion of pure magnesium, Scripta Materialia 58(2008) 421–424.

[14] N. Winzer, A. Atrens, G. Song, E. Ghali, W. Dietzel, K.U. Kainer, N. Hort, C.Blawert, A critical review of the stress corrosion cracking (SCC) of magnesiumalloys, Advanced Engineering Materials 7 (2005) 659–693.

[15] N. Winzer, A. Atrens, W. Dietzel, G. Song, K.U. Kainer, The fractography ofstress corrosion cracking (SCC) of Mg-Al alloys, Metallurgical and MaterialsTransactions A 39 (2008) 1157.

[16] J.X. Jia, G.L. Song, A. Atrens, Influence of geometry on galvanic corrosion ofAZ91D coupled to steel, Corrosion Science 48 (2006) 2133–2153.

[17] J.X. Jia, A. Atrens, G. Song, T. Muster, Simulation of galvanic corrosion ofmagnesium coupled to a steel fastener in NaCl solution, Materials andCorrosion 56 (2005) 468–474.

[18] M. Bobby Kannan, W. Dietzel, C. Blawert, A. Atrens, P. Lyon, Stress corrosioncracking of rare-earth containing magnesium alloys ZE41, QE22 and Elektron21 (EV31A) compared with AZ80, Materials Science and Engineering A 480(2008) 529–539.

[19] G.L. Song, A. Atrens, M. Dargusch, Influence of Microstructure on the corrosionof diecast AZ91D, Corrosion Science 41 (1999) 249–273.

[20] M.C. Zhao, M. Liu, G. Song, A. Atrens, Influence of homogenization annealing ofAZ91 on mechanical properties and corrosion behavior, Advanced EngineeringMaterials 10 (2008) 93–103.

[21] M.C. Zhao, M. Liu, G. Song, A. Atrens, Influence of microstructure on corrosionof as-cast ZE41, Advanced Engineering Materials 10 (2008) 104.

[22] L.M. Peng, J.W. Chang, X.W. Guo, A. Atrens, W.J. Ding, Y.H. Peng, Influence ofheat treatment and microstructure on the corrosion of magnesium alloy Mg–10Gd–3Y–0.4Zr, Journal of Applied Electrochemistry, doi:10.1007/s10800-008-9739-4.

[23] G. Song, A. Atrens, D.H. StJohn, in: J. Hryn (Ed.), Magnesium Technology, TMS,New Orleans, 2001, p. 255.

[24] G. Song, Recent progress in corrosion and protection of magnesium alloys,Advanced Engineering Materials 7 (2005) 563.

[25] G. Song, A.L. Bowles, D.H. StJohn, Corrosion resistance of aged die castmagnesium alloy AZ91D, Materials Science and Engineering A 366 (2004) 74.

[26] Z. Shi, G. Song, A. Atrens, Influence of the beta phase on the corrosionperformance of anodised coatings on magnesium-aluminium alloys, CorrosionScience 47 (2005) 2760.

[27] M.B. Haroush, C.B. Hamu, D. Eliezer, L. Wagner, The relation betweenmicrostructure and corrosion behavior of AZ80 Mg alloy following differentextrusion temperatures, Corrosion Science 50 (2008) 1766.

[28] S. Bender, J. Goellner, A. Atrens, Corrosion of AZ91 in 1N NaCl and themechanism of magnesium corrosion, Advanced Engineering Materials 10(2008) 583.

[29] G. Wu, Y. Fan, A. Atrens, C. Zhai, W. Ding, Electrochemical behavior ofmagnesium alloys AZ91D, AZCe2, and AZLa1 in chloride and sulfate solutions,Journal of Applied Electrochemistry 38 (2008) 251–257.

[30] J.W. Chang, X.W. Guo, P.H. Fu, A. Atrens, L.M. Peng, W.J. Ding, X.S. Wang, Acomparison of the corrosion behaviour in 5% NaCl solution of Mg alloys NZ30Kand AZ91D, Journal of Applied Electrochemistry 38 (2008) 207.

[31] G. Williams, H.N. McMurray, Localized corrosion of magnesium in chloride-containing electrolyte studied by a scanning vibrating electrode technique,Journal of the Electrochemical Society 155 (2008) C340–C349.

[32] W. Zhou, N.N. Aug, Y. Sun, Effect of antimony, bismuth and calcium addition oncorrosion and electrochemical behaviour of AZ91 magnesium alloy, CorrosionScience 51 (2008) 503.

[33] W.C. Neil, M. Forsyth, P.C. Howlett, C.R. Hutchinson, B.R.W. Hinton, Corrosionof magnesium alloy ZE41 - The role of microstructural feature, CorrosionScience 51 (2008) 387.

[34] M.P. Staiger, A.M. Pietak, J. Huadmai, G. Dias, Magnesium and its alloys asorthopedic biomaterials: A review, Biomaterials 27 (2006) 1728.

[35] N. Winzer, A. Atrens, W. Dietzel, G. Song, K.U. Kainer, Stress corrosion cracking(SCC) in Mg-Al alloys studied using compact specimens, Advanced EngineeringMaterials 10 (2008) 453–458.

[36] N. Winzer, A. Atrens, W. Dietzel, V.S. Raja, G. Song, K.U. Kainer,Characterisation of stress corrosion cracking (SCC) of Mg-Al alloys, MaterialsScience and Engineering A 488 (2008) 339–351.

[37] N. Winzer, A. Atrens, W. Dietzel, G. Song, K.U. Kainer, Comparison of thelinearly increasing stress test and the constant extension rate test in theevaluation of transgranular stress corrosion cracking of magnesium, MaterialsScience and Engineering A 472 (2008) 97–106.

[38] J.X. Jia, G. Song, A. Atrens, Experimental measurement and computersimulation of galvanic corrosion of magnesium coupled to steel, AdvancedEngineering Materials 9 (2007) 65–74.

[39] N. Winzer, A. Atrens, W. Dietzel, G. Song, K.U. Kainer, Evaluation of the delayedhydride cracking mechanism for transgranular stress corrosion cracking ofmagnesium alloys, Materials Science and Engineering A 466 (2007) 18–31.

[40] N. Winzer, A. Atrens, W. Dietzel, G. Song, K.U. Kainer, Transactions ofNonferrous Metals Society of China 17 (2007) S150–S155. Part A Sp. Iss. 1,NOV 2007.

[41] N. Winzer, A. Atrens, W. Dietzel, G. Song, K.U. Kainer, Stress corrosion crackingin magnesium alloys: Characterization and prevention, JOM 59 (8) (2007) 49–53.

[42] H. Meifeng, L. Lei, W. Yating, T. Zhixin, H. Wenbin, Corrosion properties ofsurface-modified AZ91D magnesium alloy, Corrosion Science 50 (2008) 3267–3273.

[43] Z. Shi, G. Song, A. Atrens, The corrosion performance of anodised magnesiumalloys, Corrosion Science 48 (2006) 3531–3546.

[44] Z. Shi, G.L. Song, A. Atrens, Influence of anodising current on the corrosionresistance of anodised AZ91D magnesium alloy, Corrosion Science 48 (2006)1939–1959.

[45] Z. Shi, G.L. Song, A. Atrens, Corrosion resistance of anodised single-phase Mgalloys, Surface and Coatings Technology 201 (2006) 492–550.

[46] A. Atrens, N. Winzer, G.L. Song, W. Dietzel, C. Blawert, Stress corrosion crackingand hydrogen diffusion in magnesium, Advanced Engineering Materials 8(2006) 749–751.

[47] R.G. Song, C. Blawert, W. Dietzel, A. Atrens, A study on stress corrosioncracking and hydrogen embrittlement of AZ31 magnesium alloy, MaterialsScience and Engineering 399 (2005) 308–317.

[48] J.X. Jia, G. Song, A. Atrens, Boundary element method predictions of theinfluence of the electrolyte on the galvanic corrosion of AZ91D coupled tosteel, Materials and Corrosion 56 (2005) 259–270.

[49] A. Atrens, Research directions in magnesium corrosion arising from theWolfsburg Conference, Advanced Engineering Materials 6 (2004) 83–84.

[50] S.B. Abhijeet, R. Balasubramaniam, M. Gupta, Corrosion behaviour of Mg-Cuand Mg-Mo composites in 3.5% NaCl, Corrosion Science 50 (2008) 2423.

[51] J. Zhang, D. Zhang, Z. Tian, J. Wang, K. Liu, H. Lu, D. Tang, J. Meng,Microstructures, tensile properties and corrosion behavior of die-cast Mg–4Al-based alloys containing La and/or Ce, Materials Science and Engineering A489 (2008) 113.

[52] E. Zhang, W. He, H. Du, K. Yang, Microstructure, mechanical properties andcorrosion properties of Mg-Zn-Y alloys with low Zn content, Materials Scienceand Engineering A 488 (2008) 102.

[53] J. Chen, J. Wang, E. Han, W. Ke, Effect of hydrogen on stress corrosion crackingof magnesium alloy in 0.1 M Na2SO4 solution, Materials Science andEngineering A 488 (2008) 428–434.

[54] M. Jönsson, D. Persson, C. Leygraf, Atmospheric corrosion of field-exposedmagnesium alloy AZ91D, Corrosion Science 50 (2008) 1406–1413.

[55] M.B. Kannan, R.K. Singh Raman, In vitro degradation and mechanical integrityof calcium-containing magnesium alloys in modified-simulated body fluid,Biomaterials 29 (2008) 2306–2314.

1292 M.-C. Zhao et al. / Corrosion Science 51 (2009) 1277–1292

[56] P. Bruesch, K. Muller, A. Atrens, H. Neff, Corrosion of stainless-steels in chloridesolution - an XPS investigation of passive films, Applied Physics A 38 (1985) 1–18.

[57] P. Bruesch, A. Atrens, K. Muller, H. Neff, Corrosion of stainless-steels in chloridesolutions – an examination of the passive films by XPS, Fresenius Zeitschriftfur Analytische Chemie 319 (1984) 812–821.

[58] P. Bruesch, A. Atrens, K. Muller, H. Neff, Corrosion of rust-free steels – an XPSstudy of passive films, Helvetica Physica Acta 57 (1984) 487.

[59] A. Froats, T.K. Aune, D. Hawke, W. Unsworth, J.E. Hillis, Metal handbook,Corrosion, 9th ed., vol. 13, ASM International Materials Park, OH, 1987, pp.740–754.

[60] Uhlig, Corrosion and Corrosion Control, Wiley, 1973.[61] Z. Xuehua, H. Yuanwei, W. Zhongling, C. Qiurong, G. Fuxing, Improvement of

corrosion resistance of AZ91D magnesium alloy by holmium addition,Corrosion Science 48 (2006) 4223.

[62] E. Zhang, L. Yang, Microstructure, mechanical properties and bio-corrosionproperties of Mg–Zn–Mn–Ca alloy for biomedical application, MaterialsScience and Engineering A 497 (2008) 111–118.

[63] T. Zhang, X. Liu, Y. Shao, G. Meng, F. Wang, Electrochemical noise analysis onthe pit corrosion susceptibility of Mg–10Gd–2Y–0.5Zr, AZ91D alloy and puremagnesium using stochastic model, Corrosion Science 50 (2008) 3500.

[64] C. Zhang, X. Huang, M. Zhang, L. Gao, R. Wu, Electrochemical characterizationof the corrosion of a Mg–Li alloy, Materials Letters 62 (2008) 2177.

[65] Samuel Brunner, Peter Richner, Ulrich Müller, Olga Guseva, Acceleratedweathering device for ser-vice life prediction of organic coating, PolymerTesting 24 (2005) 25–31.

[66] K.P. FitzGerald, J. Nairn, G. Skennerton, A. Atrens, Atmospheric corrosion ofcopper and the colour, structure and composition of natural patinas on copper,Corrosion Science 48 (2006) 2480–2509.

[67] K.P. FitzGerald, J. Nairn, A. Atrens, The chemistry of copper patination,Corrosion Science 40 (1998) 2029–2050.

[68] A. Atrens, B. Baroux, M. Mantel, The Secondary Passive Film for Type 304Stainless Steel in 0.5 M H2SO4, Journal of the Electrochemical Society 144(1997) 3697–3704.

[69] T. Nguyen, A. Atrens, Composition and morphology of the film formed on alead alloy under conditions typical of the electro-winning of copper,Hydrometallurgy 96 (2009) 14–26.

[70] T. Nguyen, A. Atrens, Influence of lead dioxide surface films on anodicoxidation of a lead alloy under conditions typical of copper electrowinning,Journal of Applied Electrochemistry 38 (2008) 569.

[71] T. Nguyen, N. Guresin, M. Nicol, A. Atrens, Influence of cobalt ions on theanodic oxidation of a lead alloy under conditions typical of copperelectrowinning, Journal of Applied Electrochemistry 38 (2008) 215.

[72] A.S. Lim, A. Atrens, ESCA studies of Cr-Co alloys, Applied Physics A 54 (1992)270–278.

[73] A.S. Lim, A. Atrens, ESCA studies of Ni-Cr alloys, Applied Physics A 54 (1992)343–349.

[74] A.S. Lim, A. Atrens, ESCA studies of Fe-Ti alloys, Applied Physics A 54 (1992)500–507.

[75] A.S. Lim, A. Atrens, ESCA studies of Si-Fe alloys, Applied Physics A 53 (1991)273–281.

[76] S. Jin, A. Atrens, Passive films on stainless-steels in aqueous-media, AppliedPhysics A 50 (1990) 287–300.

[77] A.S. Lim, A. Atrens, ESCA studies of Nitrogen-containing stainless-steels,Applied Physics A 51 (1990) 411–418.

[78] S. Jin, A. Atrens, ESCA-studies of the surface-film formed on stainless-steels byexposure to 0.1 M NaCl solution at various controlled potentials, AppliedPhysics A 46 (1988) 51–65.

[79] S. Jin, A. Atrens, ESCA-studies of the structure and composition of the passivefilm formed on stainless-steels by various immersion times in 0.1 M NaClsolution, Applied Physics A 42 (1987) 149–165.

[80] G.L. Song, A. Atrens, D.H. StJohn, J. Nairn, Y. Lang, The electrochemical corrosionof pure magnesium in 1 N NaCl, Corrosion Science 39 (1997) 855–875.

[81] G.L. Song, A. Atrens, D. StJohn, X. Wu, J. Nairn, The anodic dissolution ofmagnesium in chloride and sulphate solutions, Corrosion Science 39 (1997)1981–2004.

[82] G.L. Song, A. Atrens, X. Wu, B. Zhang, Corrosion behaviour of AZ21, AZ501 andAZ91 in sodium chloride, Corrosion Science 40 (1998) 1769–1791.

[83] G.L. Song, A.L. Bowles, D.H. StJohn, Corrosion resistance of aged die castmagnesium alloy AZ91D, Materials Science and Engineering A 366 (2004) 74.

[84] R.L. Petty, A.W. Davidson, J. Kleinberg, Journal American Chemical Society 76(1954) 363.

[85] ASM International, Metals handbook, Corrosion, ninth ed., Vol. 13, ASMInternational, 1987.